Genetic Predisposition

1 Heritability Evidence

Familial clustering and twin studies provide converging evidence that genetic factors substantially influence ME/CFS susceptibility while demonstrating that inheritance follows a complex polygenic pattern rather than simple Mendelian transmission. The gene-environment interaction model best explains the observed patterns: genetic variants establish constitutional vulnerability, but environmental triggers (particularly viral infections) are required for disease manifestation.

1.1 Twin Studies and Heritability Estimates

Twin studies offer the most rigorous method for partitioning genetic and environmental contributions to disease risk. Monozygotic (identical) twins share 100% of their DNA sequence, while dizygotic (fraternal) twins share approximately 50%. Higher concordance in monozygotic compared to dizygotic twins indicates genetic contribution, with the magnitude of the difference allowing estimation of heritability.

Buchwald et al. conducted a population-based twin study using the University of Washington Twin Registry, identifying twin pairs where at least one twin met criteria for chronic fatigue. The study found significantly higher concordance in monozygotic twins (55%) compared to dizygotic twins (19%) for unexplained chronic fatigue, yielding a heritability estimate of approximately $h^2 = 0.51$ (95% CI: 0.37–0.65). When applying more stringent CFS case definitions, monozygotic concordance decreased to approximately 30–40%, but still exceeded dizygotic concordance, suggesting heritability estimates in the range $h^2 = 0.3$–$ 0.5$ depending on phenotype definition (Buchwald et al. 2001).

These moderate heritability estimates indicate that genetic factors explain 30–50% of liability to ME/CFS, with environmental factors and gene-environment interactions accounting for the remainder. The incomplete concordance even in monozygotic twins (55% rather than 100%) demonstrates that genetic susceptibility alone is insufficient for disease development. Australian twin registry studies corroborate these findings, with concordance patterns consistent with polygenic inheritance and substantial environmental contribution. The moderate heritability is similar to other complex diseases such as type 2 diabetes ($h^2 \approx 0.4$–$ 0.6$) and autoimmune conditions, supporting classification of ME/CFS as a multifactorial disorder arising from interactions between multiple genetic variants and environmental exposures.

1.2 Familial Aggregation and Relative Risk

Family studies complement twin research by examining disease clustering across multiple generations and family structures. Multiple independent studies document elevated ME/CFS prevalence among first-degree relatives of affected individuals compared to the general population.

Walsh et al. conducted a family study examining relatives of ME/CFS probands and found that first-degree relatives had significantly increased risk, though precise relative risk estimates varied depending on diagnostic criteria and ascertainment methods (Walsh et al. 2001). The observed familial aggregation persisted after controlling for shared household exposures during childhood, arguing against purely environmental transmission through common viral exposures or psychosocial factors. Families with multiple affected members often show variable clinical presentations, suggesting that shared genetic susceptibility manifests differently depending on individual trigger exposures and additional genetic modifiers.

The pattern of familial clustering shows several notable features. First, affected relatives frequently report different precipitating events (different viral infections, surgeries, traumas), indicating that the inherited component reflects general vulnerability rather than specific pathogen susceptibility. Second, age of onset varies widely among affected family members, suggesting that the genetic liability does not determine timing but rather establishes lifelong susceptibility that may be triggered at any point. Third, affected relatives may show different dominant symptom profiles (some predominantly immunological, others metabolic or neurological), consistent with the hypothesis that shared genetic factors establish broad systemic vulnerability that interacts with individual-specific factors to determine phenotypic expression.

1.3 Gene-Environment Interaction Model

The gene-environment interaction framework provides the most parsimonious explanation for observed inheritance patterns. Genetic variants establish constitutional susceptibility, but environmental triggers are necessary and often sufficient to precipitate disease in genetically vulnerable individuals. This model explains several key observations that pure genetic or pure environmental models cannot.

The NIH RECOVER study found that 4.5% of COVID-19 survivors developed ME/CFS (Komaroff and Lipkin 2023), meaning 95.5% recovered fully despite identical viral exposure. This dramatic variation in outcome following a common environmental trigger strongly implicates genetic factors in determining who progresses from acute infection to chronic illness. Similarly, the well-documented association between infectious mononucleosis and subsequent ME/CFS affects only a minority of those infected with Epstein-Barr virus, despite near-universal infection rates by adulthood in most populations. Giardia lamblia outbreaks provide natural experiments: following the 2004 Bergen, Norway outbreak, approximately 5% of exposed individuals developed chronic fatigue meeting ME/CFS criteria, while 95% recovered completely, again demonstrating genetic influence on chronic sequelae following identical pathogen exposure.

Children of ME/CFS parents inherit elevated risk compared to the general population, but most do not develop the condition. This pattern reflects the polygenic architecture: multiple risk variants segregate through families, with children inheriting various combinations. Some inherit many risk alleles and show high genetic liability, others inherit few and have risk approaching population baseline. Environmental trigger exposure then interacts with this inherited liability: children with high genetic loading may develop ME/CFS following relatively mild infections, while those with low genetic loading may remain unaffected even after severe viral illnesses. Intermediate genetic liability creates vulnerability to severe triggers but resilience against mild ones.

The specific genetic variants inherited may influence which environmental triggers are most pathogenic. For example, children inheriting immune gene variants affecting viral immune responses may be particularly susceptible to viral triggers but less susceptible to non-infectious stressors. Those inheriting metabolic gene variants might be vulnerable to physical or metabolic stressors. This genotype-specific susceptibility to different environmental factors could explain the heterogeneity in precipitating events observed even within affected families.

2 Genetic Variants and Candidate Gene Studies

Before the advent of genome-wide association studies, candidate gene approaches investigated single nucleotide polymorphisms (SNPs) in genes hypothesized to influence ME/CFS pathophysiology. These studies focused on immune system genes, metabolic pathways, neurotransmitter systems, and stress response mechanisms. While candidate gene studies have important methodological limitations—including small sample sizes, multiple testing issues, and publication bias—they have identified several plausible genetic associations that warrant further investigation in larger cohorts.

2.1 Human Leukocyte Antigen (HLA) Associations

The HLA complex on chromosome 6p21 encodes major histocompatibility complex (MHC) proteins that present antigens to T cells and play a central role in adaptive immunity. HLA alleles show strong associations with autoimmune diseases, and several studies have examined whether specific HLA types predispose to ME/CFS.

Multiple studies have reported associations between ME/CFS and specific HLA class II alleles, particularly HLA-DRB1 and HLA-DQ variants. Several studies have identified increased frequency of specific HLA-DQA1 alleles in ME/CFS patients compared to controls, suggesting a potential role for antigen presentation in disease pathogenesis. Carlo-Stella et al. found associations with HLA-DQ3, particularly in patients with post-infectious onset (Carlo-Stella et al. 2009). However, these associations have shown inconsistent replication across populations, likely reflecting both genuine population differences in HLA allele frequencies and the polygenic architecture of ME/CFS where HLA contributes modest effect size as one of many susceptibility loci.

The biological plausibility of HLA associations is strong. HLA molecules determine which viral and self-peptides are presented to T cells, influencing both antiviral immune responses and potential autoreactivity. Specific HLA alleles might predispose to inefficient viral clearance, prolonged antigenic stimulation, or molecular mimicry leading to autoimmune sequelae. The connection to post-infectious onset supports this mechanism: individuals with particular HLA types may mount ineffective immune responses to triggering infections, permitting viral persistence or chronic immune activation.

2.2 Immune System Gene Variants

Beyond HLA, numerous genes regulating innate and adaptive immunity have been examined as ME/CFS susceptibility loci.

Cytokine and cytokine receptor genes represent logical candidates given the well-documented cytokine dysregulation in ME/CFS (Chapter Immune System Dysfunction). Polymorphisms in TNF-$\alpha$ promoter region (particularly the $-308$ G/A variant associated with higher TNF-$\alpha$ production) have been investigated, with some studies reporting increased frequency of high-expression alleles in ME/CFS patients. Variants in IL-10 (an anti-inflammatory cytokine), IL-6, and IL-1 gene clusters have also been examined. Goertzel et al. reported associations with variants affecting IL-10 expression, consistent with the hypothesis that impaired anti-inflammatory responses permit chronic inflammation (Goertzel et al. 2006).

Toll-like receptor (TLR) genes, which recognize pathogen-associated molecular patterns and initiate innate immune responses, have shown suggestive associations. TLR4 polymorphisms affecting responsiveness to bacterial lipopolysaccharide may influence susceptibility to post-infectious ME/CFS. Pattern recognition receptor variants could plausibly affect both initial pathogen detection and subsequent inflammatory cascades.

Complement system genes have received less attention but merit investigation given emerging evidence for complement dysregulation in ME/CFS. Genetic variants affecting complement activation thresholds or regulatory protein function might predispose to excessive inflammatory responses or impaired clearance of immune complexes.

DNASE1 and DNASE1L3 — Potential Post-Viral Susceptibility Loci. Garcia et al. (Garcia et al. 2024) identified nine linked DNASE1 promoter polymorphisms associated with ~75% reduction in DNase1 antigen in COVID-19 patients, all three minor-allele carriers being critical cases. These variants reduce circulating DNase1—one of two enzymes responsible for degrading neutrophil extracellular traps (NETs)—creating a constitutional vulnerability to NET/DNase imbalance during severe infection. If post-infectious ME/CFS involves impaired NET clearance (Chapter Immune System Dysfunction), then DNASE1 loss-of-function variants could represent genetic susceptibility factors predisposing to chronic post-viral pathology. DNASE1L3 variants did not show significant associations in this cohort, though sample size ($n=52$ hospitalized) limited power for rare variant detection. No DNASE1 or DNASE1L3 genotyping has been performed in ME/CFS cohorts.

Certainty: 0.15 (ME/CFS extrapolation; variant-phenotype association established in COVID-19 only; no ME/CFS-specific genotyping data; mechanism plausibility moderate).

2.3 Metabolic and Mitochondrial Gene Variants

The profound metabolic dysfunction documented in ME/CFS (Chapter Energy Metabolism and Mitochondrial Function) suggests that genetic variants affecting cellular energetics may contribute to disease susceptibility.

Mitochondrial DNA (mtDNA) variants have been examined in several studies, though results remain inconclusive. Unlike nuclear DNA, mtDNA is maternally inherited and shows high mutation rates. Some studies have reported increased mtDNA deletions or specific haplogroup associations in ME/CFS, but replication has been inconsistent. The biological rationale remains strong: mtDNA encodes critical components of the electron transport chain, and variants reducing mitochondrial efficiency could predispose to bioenergetic crisis under conditions of increased demand or oxidative stress.

Nuclear genes encoding mitochondrial proteins represent equally plausible candidates. Recent evidence identifies WASF3 pathway dysregulation in ME/CFS (Syed et al. 2025), potentially affecting cellular energy production capacity. WASF3 is involved in actin cytoskeleton regulation and mitochondrial dynamics; genetic variants affecting its expression or function might impair mitochondrial quality control mechanisms or cellular energy distribution.

Genes involved in glucose metabolism, fatty acid oxidation, and oxidative stress responses have shown suggestive associations in small studies. Polymorphisms affecting glycolytic enzyme expression, carnitine transport (relevant for fatty acid metabolism), or antioxidant systems (superoxide dismutase, catalase, glutathione pathways) could plausibly influence metabolic reserve and stress tolerance.

The C677T polymorphism (rs1801133) in the methylenetetrahydrofolate reductase (MTHFR) gene is among the most common functional single-nucleotide polymorphisms in the human genome and reduces MTHFR enzyme activity to approximately 67% in heterozygotes and 25% in TT homozygotes (Zarembska, Ślusarczyk, and Wrzosek 2023). This impairs conversion of folate to 5-methyltetrahydrofolate, the methyl donor required for homocysteine remethylation to methionine. Consequently, C677T carriers are predisposed to elevated homocysteine under conditions of folate or B12 insufficiency.

Disrupted one-carbon metabolism from MTHFR dysfunction affects not only cytosolic methylation but also mitochondrial DNA methylation, with potential downstream effects on oxidative phosphorylation (OXPHOS) gene expression (Zarembska, Ślusarczyk, and Wrzosek 2023). A single case report illustrates the potential clinical extreme of this pathway: an adolescent with compound heterozygous MTHFR mutations, markedly elevated serum homocysteine (86 $\mu$mol/L), and comorbid ME/CFS, postural tachycardia syndrome (POTS), and narcolepsy (Liao et al. 2021) (note: n=1; cannot establish causality).

The Regland 2015 study found that folate dosing matched to MTHFR genotype was associated with better B12/folate treatment response in ME/FM patients, suggesting pharmacogenomic relevance (Regland et al. 2015).

(Certainty: Medium-High for biochemistry; Low for ME/CFS-specific association.)

Open Question: MTHFR Prevalence in ME/CFS

Whether the MTHFR C677T TT genotype is overrepresented in ME/CFS relative to the general population has not been established by appropriately powered genetic association studies.

2.4 Ion Channel and Neurotransmitter System Genes

Neurological symptoms in ME/CFS (Chapter Neurological and Neurocognitive Dysfunction) and the documented dysfunction of transient receptor potential melastatin 3 (TRPM3) ion channels suggest genetic variants in ion channel genes as susceptibility factors.

TRPM3 dysfunction represents one of the most mechanistically informative genetic associations. Marshall-Gradisnik and colleagues have demonstrated reduced TRPM3 function in ME/CFS patients’ natural killer cells and B cells, with impaired calcium influx following TRPM3 activation (Marshall-Gradisnik et al. 2016). While these functional studies demonstrate acquired TRPM3 dysfunction, genetic variants in the TRPM3 gene (particularly regulatory variants affecting expression levels) could establish constitutional vulnerability. TRPM3 channels regulate calcium signaling, which is essential for immune cell function, neurotransmitter release, and cellular metabolism. Reduced baseline TRPM3 expression due to genetic variants might create a narrower functional reserve, rendering individuals more susceptible to further acquired dysfunction.

Other ion channel genes merit investigation. Voltage-gated calcium channels, potassium channels regulating neuronal excitability, and acid-sensing ion channels (ASICs) involved in pain perception and autonomic regulation all represent plausible candidates. Channelopathies—diseases caused by ion channel dysfunction—often present with episodic symptoms, fatigue, and autonomic features resembling aspects of ME/CFS.

Neurotransmitter system genes, particularly those affecting serotonin, norepinephrine, and dopamine metabolism, have been examined given the prominent cognitive and mood symptoms. The catechol-O-methyltransferase (COMT) gene, which catabolizes catecholamines, exists in high-activity (Val158) and low-activity (Met158) variants. Some studies have reported associations with the Met158 variant, which would reduce catecholamine degradation and potentially affect stress responses and cognitive function. Serotonin transporter (5-HTTLPR) polymorphisms affecting serotonin reuptake have shown inconsistent associations.

2.5 Autonomic and Cardiovascular Genes

The high prevalence of orthostatic intolerance and postural orthostatic tachycardia syndrome (POTS) in ME/CFS patients (co-occurring in approximately 60%) (Natelson et al. 2022) suggests genetic overlap with autonomic dysfunction syndromes.

Adrenergic receptor genes, particularly $\beta$-adrenergic receptor variants affecting cardiac responsiveness to catecholamines, represent logical candidates. The $\beta_1$-adrenergic receptor gene (ADRB1) shows common polymorphisms affecting receptor density and signaling efficiency. Variants that alter cardiovascular responsiveness to sympathetic activation could predispose to orthostatic intolerance, particularly when combined with other ME/CFS-related pathophysiology such as reduced blood volume or impaired baroreceptor function.

Genes affecting renin-angiotensin-aldosterone system (RAAS) function, which regulates blood volume and vascular tone, could influence susceptibility to orthostatic symptoms. ACE (angiotensin-converting enzyme) gene variants, particularly the insertion/deletion polymorphism affecting ACE levels, might interact with other cardiovascular genetic factors to determine orthostatic tolerance.

2.6 Limitations of Candidate Gene Studies

Warning: Candidate Gene Study Limitations

Most candidate gene studies in ME/CFS suffer from serious methodological limitations that prevent definitive conclusions. Common issues include small sample sizes (often n < 100 cases), which provide insufficient statistical power to detect modest genetic effects; inadequate correction for multiple testing, leading to false positive findings; publication bias favoring positive associations; and lack of independent replication in separate cohorts. Many reported associations have not been replicated, and effect sizes when reported are often implausibly large, suggesting winner’s curse (overestimation of effect size in discovery samples).

The transition to genome-wide association studies addresses many of these limitations through systematic interrogation of common genetic variation across the entire genome, adequate sample sizes to detect realistic effect sizes, stringent correction for multiple testing, and consortia-based designs facilitating replication.

3 Genome-Wide Association Studies

Genome-wide association studies (GWAS) represent a paradigm shift from candidate gene approaches, systematically interrogating millions of common genetic variants across the entire genome to identify disease-associated loci without prior hypotheses about specific genes. GWAS have successfully identified genetic risk factors for numerous complex diseases including type 2 diabetes, inflammatory bowel disease, schizophrenia, and rheumatoid arthritis. For ME/CFS, GWAS has been hindered by the challenges of patient recruitment, diagnostic heterogeneity, and the need for large sample sizes to detect the modest effect sizes typical of complex polygenic diseases.

3.1 DecodeME: The Largest ME/CFS GWAS

Achievement: DecodeME GWAS Findings

The DecodeME study represents the largest genetic investigation of ME/CFS to date, with over 15,000 ME/CFS patients contributing DNA samples, compared against population controls through the UK Biobank and other cohorts (DecodeME Consortium, Ponting, et al. 2025). This large sample size provides statistical power to detect genetic variants with realistic effect sizes (odds ratios of 1.1–1.3) that reach genome-wide significance ($p < 5 \times 10^{-8}$).

DecodeME employed rigorous case ascertainment through physician diagnosis and self-report with verification, accepting patients meeting CCC (Canadian Consensus Criteria), ICC (International Consensus Criteria), or IOM (Institute of Medicine) diagnostic criteria. This inclusive approach maximizes sample size while acknowledging diagnostic heterogeneity, with sensitivity analyses examining whether genetic architecture differs across diagnostic subtypes.

The study’s scale enables several key analyses beyond simple case-control association: estimation of SNP heritability (the proportion of ME/CFS liability explained by common genetic variants), genetic correlation analyses comparing ME/CFS to other conditions, polygenic risk score development, and gene-based and pathway enrichment tests identifying biological systems enriched for associated variants.

Warning: Replication Status: Not Yet Replicated (By Design)

DecodeME is the first adequately powered ME/CFS GWAS. The genome-wide significant loci identified await replication by design—GWAS discovery cohorts require independent replication cohorts to confirm true associations versus false positives. No independent replication cohort of comparable size currently exists, though international collaborations (e.g., with US ME/CFS Clinician Coalition biobanks) may provide future replication opportunities.

3.1.1 SNP Heritability

DecodeME estimated SNP-based heritability at $h^2_\text{SNP} = 0.095$ (9.5%) on the liability scale via LD Score Regression (DecodeME Consortium, Ponting, et al. 2025) (ME/CFS Science 2025a). This is modest compared to schizophrenia ($h^2_\text{SNP} \approx 0.26$), Crohn’s disease ($\approx 0.24$), or type 1 diabetes ($\approx 0.22$), but close to the mean heritability across all UK Biobank traits ($h^2 \approx 0.10$) and substantially higher than fibromyalgia ($h^2_\text{SNP} \approx 0.01$). The gap between twin-study heritability ($h^2 \approx 0.3$–$ 0.5$; Section Genetic Predisposition) and SNP heritability reflects the “missing heritability” expected for complex traits: rare variants, structural variants, gene–gene interactions, and epigenetic contributions are not captured by common-variant GWAS arrays.

3.1.2 The Eight Genome-Wide Significant Loci

DecodeME identified eight loci reaching genome-wide significance ($p < 5 \times 10^{-8}$) across more than 8 million tested SNPs (DecodeME Consortium, Ponting, et al. 2025) (ME/CFS Science 2025a). [^1] All eight are common variants (minor allele frequency 13–63%) with small individual effect sizes (odds ratios 0.93–1.10), consistent with the polygenic architecture expected for complex diseases. The prevalence difference between patients and controls at each locus is only 1–2 percentage points—for example, the chr17 lead SNP is present in 34.7% of ME/CFS cases versus 32.9% of controls (ME/CFS Science 2025a). These eight loci represent the tip of a polygenic iceberg: hundreds or thousands of additional variants likely contribute sub-threshold effects.

The strongest signal by far is on chromosome 20 ($-log_10 p = 11.02$, OR $= 1.095$) in a region containing three candidate genes: ARFGEF2 (vesicle trafficking), CSE1L (nuclear transport), and STAU1 (mRNA transport)—all highlighted by MAGMA gene-based testing (ME/CFS Science 2025b). Other genome-wide significant loci implicate CA10 (chr17; CNS development, chronic pain), UNC13C (chr15; glutamatergic synaptic transmission), and OLFM4 (chr13; neural development, innate immunity) (DecodeME Consortium, Ponting, et al. 2025) (ME/CFS Science 2025b).

A critical interpretive point: small GWAS effect sizes do not predict therapeutic utility. As analysts at mecfsscience.org note, GWAS effect sizes have not predicted drug development success—HMGCR variants have tiny GWAS effects, yet statins (HMGCR inhibitors) produce substantial clinical benefit (ME/CFS Science 2025a) (King, Davis, and Degner 2024). That said, HMGCR is the exception rather than the rule: more than 90% of GWAS-nominated drug targets fail in clinical development, so small genetic effect sizes remain a weak basis for therapeutic optimism without additional functional validation. Moreover, because DNA is fixed at birth, genetic associations are not confounded by illness-related changes (deconditioning, medication use, sleep disruption), making them stronger evidence for causal pathways than most ME/CFS biomarker studies.

3.1.3 Gene Prioritisation and Fine-Mapping

Most GWAS signals lie in non-coding regions, necessitating gene prioritisation through expression quantitative trait locus (eQTL) mapping, tissue expression analysis, and proximity-based inference (ME/CFS Science 2025b). DecodeME used Genotype-Tissue Expression (GTEx) project data to correlate lead SNPs with gene expression across tissues. A key limitation is that only approximately 40% of GWAS signals can be matched to specific genes via eQTL approaches.

Achievement: DecodeME Brain Tissue Enrichment

Certainty: 0.80. DecodeME’s MAGMA gene-tissue analysis revealed significant enrichment of ME/CFS-related genes in all 13 brain tissues examined, providing the first genetic evidence for central nervous system involvement in ME/CFS (DecodeME Consortium, Ponting, et al. 2025). This brain-wide enrichment pattern suggests that ME/CFS genetic susceptibility operates through neural pathways rather than peripheral immune or metabolic mechanisms alone. Implicated genes include CA10 (CNS development, chronic pain), ARFGEF2 (vesicle trafficking, neuronal function), and UNC13C (glutamatergic synaptic transmission). The brain tissue enrichment finding is consistent with the neuroimaging abnormalities, neurochemical dysregulation, and HPA axis dysfunction documented in ME/CFS (Chapters Neurological and Neurocognitive Dysfunction and Endocrine and Metabolic Dysfunction).

This finding has been replicated and extended by the Maccallini 2026 meta-GWAS (n=19,470 cases), which found exclusive brain and pituitary enrichment across 30 tissues and additionally performed cell-type enrichment identifying medium spiny neurons (MSNs) in the striatum as the most specific cell-type hit (Maccallini 2026). Stratified LDSC analysis using the Finucane et al. 2018 pipeline replicated the CNS signal across multiple reference datasets (GTEx v6, Dropviz mouse atlas, Human Brain Atlas) and found zero significant immune cell-type associations (Finucane et al. 2018). A third independent analysis using the DESCARTES fetal human atlas confirmed neuronal rather than immune cell-type involvement (Lee et al. 2026). See Achievement Cell-Type Enrichment Analyses Converge on Neuronal Signal in Chapter Neurological and Neurocognitive Dysfunction for the cell-type specificity discussion.

Study: (n=15,579 cases, n=259,909 controls, MAGMA analysis, 13 brain tissues, replicated in meta-GWAS n=19,470, convergent with 3 cell atlases; key caveat: fine cell-type resolution is method-dependent).

The MZ twin discordant study proposal in Chapter 47 (MZ Twin Discordant Design: Striatal Imaging, Microbiome, and LSR in Genetically Controlled ME/CFS) provides a genetically controlled design to test whether this neuronal genetic signal translates to functional striatal pathology.

Fine-mapping of the eight significant loci plus sub-threshold hits ($p < 5 \times 10^{-7}$) identifies three convergent biological themes (ME/CFS Science 2025b):

Neuronal development and synaptic function. The clearest and most consistent signal across DecodeME loci points to the brain. MAGMA gene-set analysis found that all significantly enriched tissues were brain regions (ME/CFS Science 2025a). Specific genes include:

CA10 (chr17): role in CNS development; previously associated with insomnia, chronic pain, and restless legs syndrome
SHISA6 (chr17, sub-threshold): excitatory synaptic transmission at glutamatergic synapses; associated with sleep duration and insomnia
SOX6 (chr11, sub-threshold): required for normal CNS development
LRRC7 (chr1, sub-threshold): regulation of neuron projection development; associated with depression, educational attainment
DCC (chr18, sub-threshold): mediates axon guidance of neuronal growth cones; associated with insomnia, pain, depression, autism, schizophrenia
UNC13C (chr15): predicted involvement in glutamatergic synaptic transmission
BARHL2 (chr1, sub-threshold): neuron generation and axon extension

The convergence of these genes on neuronal communication and synaptic function is notable. Cross-referencing with the GWAS Catalog reveals that several of these genes have been independently associated with insomnia, depression, and chronic pain in other GWAS—conditions that share substantial clinical overlap with ME/CFS (ME/CFS Science 2025b). According to the mecfsscience.org analysis, OLFM4 and DCC were also identified in a 2025 fibromyalgia GWAS preprint, though the specific DNA signals at these loci differ between the two conditions, suggesting the same genes may be involved through different regulatory mechanisms (ME/CFS Science 2025b). This overlap is now confirmed by the published multi-ancestry fibromyalgia GWAS of Kerrebijn et al. (2026), which reports fibromyalgia risk regions overlapping with ME/CFS at OLFM4 and RABGAP1L (DCC was flagged in the earlier overlap), while GPR52/HTT are fibromyalgia-risk genes whose presence in ME/CFS genetics remains untested (Section Three-Line Genetic Convergence on Neuronal Biology) (Kerrebijn et al. 2026).

Immune function (ambiguous). Several loci contain immune-relevant genes, but gene-dense regions create assignment ambiguity (ME/CFS Science 2025b):

RABGAP1L (chr1): involved in bacterial expulsion from cells and limiting viral replication; however, this locus contains 11 candidate genes
BTN2A2 (chr6): butyrophilin family gene in the MHC region, packed with immune genes
TAOK3 (chr12, proximity-only inference; not in DecodeME’s prioritised gene list): role in T-cell activation, linked to lupus; lies closest to the chr12 signal
HLA-DQA1*05:01: associated with ME/CFS at genome-wide significance, but the HLA region is notoriously difficult to fine-map; further analyses planned

Autophagy and intracellular transport. An unexpected signal emerged around cellular quality control mechanisms (ME/CFS Science 2025b):

FBXL4 (chr6, sub-threshold; gene assignment not definitive): involved in mitophagy (selective autophagy of mitochondria); La Trobe University plans further exploration
CCPG1 (chr15): facilitates ER-phagy (endoplasmic reticulum autophagy)
ARFGEF2, CSE1L, STAU1 (chr20, the strongest locus): all involved in vesicle trafficking and intracellular transport

The autophagy signal is intriguing given the well-documented mitochondrial dysfunction in ME/CFS (Chapter Energy Metabolism and Mitochondrial Function). A possible mechanistic connection—that genetically impaired mitophagy could permit accumulation of damaged mitochondria, contributing to bioenergetic crisis—is formalised as Hypothesis Genetic Mitophagy Vulnerability: The Accumulation Threshold Model below.

3.1.4 Convergence with Rare-Variant Studies

As reported by mecfsscience.org, the Stanford Mark Snyder laboratory independently published a preprint (2025) using a fundamentally different approach: instead of common variants, they studied rare loss-of-function variants using a neural network trained on biological data (ME/CFS Science 2025b). Their risk genes—including NLGN2 and SYNGAP1—are reported to be involved in synaptic function, aligning with the DecodeME common-variant findings. If confirmed, this convergence of two independent methodologies on neuronal communication genes would substantially strengthen the case that brain-related genetic pathways contribute to ME/CFS susceptibility. The planned SequenceME study will specifically target rare variants (minor allele frequency $< 1%$), which may show larger, clearer effects than the common variants captured by DecodeME.

3.1.5 Confounding and Robustness

DecodeME restricted analysis to British participants with European ancestry and corrected for population stratification using PCA with the first 20 principal components, achieving a genomic inflation factor of 1.066 (close to the ideal 1.0), indicating minimal confounding (ME/CFS Science 2025b). A potential concern is symptom confounding: several implicated genes have prior associations with depression, insomnia, and pain. These associations could reflect genuine shared biology (common neural pathways disrupted in ME/CFS and these conditions) or confounding if control groups contain undiagnosed individuals with these symptoms. The DecodeME team plans future analyses testing whether associations hold in ME/CFS patients without comorbid depression, using the study’s detailed questionnaire data.

3.2 Genetic Correlations with Other Conditions

DecodeME computed genetic correlations ($r_g$) via LD Score Regression against 3,167 traits in the UK Biobank BIGA database (DecodeME Consortium, Ponting, et al. 2025) (ME/CFS Science 2025a). The results are notable for what they include and what they exclude.

The strongest genetic correlations (all Bonferroni-corrected for 3,167 comparisons; threshold $\approx 1.6 \times 10^{-5}$) are with functional and symptom-based conditions:

Irritable bowel syndrome: $r_g = 0.75$ ($p = 0.00015$); confirmed by twin-sibling and family studies showing shared genetic architecture between CFS, fibromyalgia, and IBS (Steen et al. 2026) (Kendler et al. 2023)
Self-reported chronic fatigue syndrome: $r_g = 0.70$ ($p = 0.00005$)
Sleeping too much: $r_g = 0.66$ ($p = 0.00028$)
Depression (professionally diagnosed): $r_g = 0.60$ ($p < 0.00001$)
Amitriptyline use: $r_g = 0.61$ ($p < 0.00001$)
Recent tiredness/low energy: $r_g = 0.61$ ($p < 0.00001$)
Spondylosis: $r_g = 0.59$ ($p = 0.016$; nominally significant but does not survive strict Bonferroni correction)

The high correlation with IBS ($r_g = 0.75$) is consistent with the well-documented clinical overlap and may reflect shared autonomic, gut–brain axis, or mast cell pathways. The depression correlation ($r_g = 0.60$) does not imply that ME/CFS is a depressive disorder; rather, shared genetic variants likely affect common neurological substrates (sleep regulation, pain processing, fatigue signaling) that are disrupted in both conditions through different downstream mechanisms. Childhood asthma shows a modest correlation ($r_g = 0.31$; p-value not reported as Bonferroni-significant), consistent with immune pathway involvement (ME/CFS Science 2025a).

Limitation: Genetic Correlation $\neq$ Shared Aetiology

Genetic correlation measures shared common-variant architecture, not shared causal mechanisms. Two conditions can show high $r_g$ because they share upstream risk variants that diverge into different downstream pathologies. The $r_g = 0.60$ with depression, for example, may reflect that neuronal genes affecting synaptic function (such as those identified at DecodeME loci) contribute to both conditions through different circuits. Critically, ME/CFS shows no significant genetic correlation with classic autoimmune diseases: multiple sclerosis, rheumatoid arthritis, Crohn’s disease, type 1 diabetes, or type 2 diabetes (ME/CFS Science 2025a). This absence challenges hypotheses positioning ME/CFS as primarily an autoimmune condition and suggests that any autoimmune component operates through mechanisms genetically distinct from established autoimmune diseases.

Speculation: Depression Paradox: High r_g, No Shared Causal Variants

The high genetic correlation with depression ($r_g = 0.60$, $p < 0.00001$) has been interpreted by some as evidence that ME/CFS is fundamentally a mood disorder. DecodeME’s finding of no shared causal variants between ME/CFS and depression resolves this apparent paradox: glutamatergic synaptic genes are shared risk factors, but circuit expression differs — prefrontal-limbic circuits in depression versus cortico-cerebellar and brainstem circuits in ME/CFS (Maccallini 2026) (DecodeME Consortium, Ponting, et al. 2025). This circuit-level divergence explains how the same genetic variants can produce clinically distinct conditions. The depression correlation is therefore a signal of shared synaptic biology, not shared psychiatric aetiology.

(Certainty: 0.50)

Falsifiable prediction: Brain-region-specific eQTL analysis will demonstrate that ME/CFS-associated glutamatergic variants have differential expression effects between cerebellum/brainstem and prefrontal cortex, while depression-associated variants show the opposite regional pattern (prefrontal > cerebellar). If glutamatergic variants show equivalent expression effects in both regional profiles, circuit-level divergence fails as an explanation for the r_g/causal-variant paradox.

3.3 Polygenic Risk Scores

Polygenic risk scores (PRS) aggregate the effects of thousands or millions of genetic variants into a single quantitative measure of inherited liability. PRS can identify individuals at high genetic risk (top decile of PRS distribution), who may benefit from preventive interventions, or individuals at low genetic risk despite environmental exposures. For ME/CFS, PRS applications include risk stratification, mechanistic subtyping, and prediction.

DecodeME could enable development of ME/CFS polygenic risk scores, though none has yet been tested for clinical utility.

Open Question: ME/CFS Polygenic Risk Score Clinical Utility

Beyond individual risk prediction, PRS may also resolve the longstanding subtype problem in ME/CFS. If patients with high neuronal-gene PRS differ clinically from those with high autophagy-gene or immune-gene PRS, this would enable biologically coherent subtyping that could transform clinical trial design from “one size fits all” to pharmacogenomic stratification (see Open Question Genetic Subtypes in ME/CFS, Chapter Integrative Models and Multi-System Pathophysiology, and proposed study design in Chapter Entries added 2026-08-26: Central Motor-Drive Fatigability Cascade (Bedard 2026)). The Maccallini 2026 meta-GWAS, with nearly 20,000 cases, substantially expands the discovery sample for PRS development and may improve predictive performance through better effect-size estimation (Maccallini 2026). Several key questions remain untested: Does high PRS predict which individuals develop ME/CFS following infectious mononucleosis or COVID-19? Do patients with high versus low PRS show different clinical phenotypes, treatment responses, or prognoses? Can PRS combined with environmental risk factors improve prediction compared to either alone? The clinical utility of PRS depends on effect size distribution. If ME/CFS liability reflects thousands of variants each contributing tiny effects, PRS discriminative ability may be modest (for illustration, AUC approximately 0.6–0.65), limiting clinical utility. If a subset of variants have larger effects, PRS performance improves. Even modest predictive ability could have clinical value if the risk gradient between high and low PRS is sufficiently large to guide post-exposure monitoring and early intervention.

3.4 Earlier GWAS Attempts and Methodological Challenges

Prior to DecodeME, several smaller GWAS attempts were conducted with sample sizes of 200–500 cases. These studies were severely underpowered to detect realistic effect sizes for complex disease variants and produced no genome-wide significant findings that replicated. This failure reflects general principles of GWAS: detecting odds ratios of 1.1–1.2 (typical for complex disease variants) requires thousands to tens of thousands of cases, not hundreds. Notably, Ueland et al. (2022) explicitly attempted to replicate the TRA locus association from earlier candidate studies and found no significant association in their Norwegian cohort (Ueland et al. 2022), a null result that reflects the field’s maturation from candidate-gene to GWAS-driven approaches.

Small GWAS can still provide value through polygenic analyses aggregating information across many sub-threshold variants and through contributing data to meta-analyses. However, their inability to identify genome-wide significant loci frustrated early genetic investigation of ME/CFS and highlighted the necessity of large collaborative efforts. Several methodological challenges complicate ME/CFS GWAS beyond simply achieving adequate sample size. Diagnostic heterogeneity creates noise: if different diagnostic criteria capture partially overlapping patient populations with different genetic architectures, this heterogeneity reduces power. Potential solutions include stratified analyses by diagnostic criteria and phenotype refinement using quantitative traits (severity scores, specific symptoms) rather than binary case-control status.

Population stratification—systematic ancestry differences between cases and controls—can produce spurious associations. Standard GWAS methods correct for stratification using principal components analysis of genetic data, ensuring cases and controls are matched for genetic ancestry. For ME/CFS, international collaborative GWAS must carefully model ancestry structure to avoid confounding.

The missing heritability problem—the gap between twin study heritability estimates and SNP heritability from GWAS—arises from several sources. Rare variants (minor allele frequency $< 1%$) not well captured by standard GWAS arrays may contribute to liability. Structural variants, copy number variations, and epigenetic modifications are not directly tested in GWAS. Gene-gene and gene-environment interactions may contribute to liability but are difficult to detect with current methods. Nevertheless, GWAS SNP heritability typically explains 20–50% of twin study heritability for complex diseases, providing genome-wide validation of genetic contribution while highlighting areas for future investigation.

3.5 Implications for Understanding ME/CFS Pathophysiology

DecodeME’s findings reshape understanding of ME/CFS genetic architecture in three ways. First, the overwhelming convergence on brain-expressed genes—confirmed by both MAGMA tissue enrichment and individual gene characterisation—suggests neuronal dysfunction as a genetically grounded component of ME/CFS, rather than merely a downstream consequence of peripheral pathology (DecodeME Consortium, Ponting, et al. 2025) (ME/CFS Science 2025a) (ME/CFS Science 2025b). The glutamatergic synapse genes (SHISA6, UNC13C) are particularly noteworthy given the neurological abnormalities described in Chapter Neurological and Neurocognitive Dysfunction, where altered neurotransmitter dynamics have been observed.

Second, the absence of genetic correlation with established autoimmune diseases constrains the autoimmune hypothesis: if ME/CFS involved the same genetic liability as MS, RA, or T1D, this would appear as significant $r_g$. The autoimmune features documented in Chapter Immune System Dysfunction may therefore represent a phenocopy driven by different genetic pathways—perhaps the immune-ambiguous loci identified in DecodeME (Section Replication Status: Not Yet Replicated (By Design)) operating through novel mechanisms rather than classical autoimmunity.

Third, the autophagy/mitophagy genes (FBXL4, CCPG1) provide a genetic link to the mitochondrial dysfunction documented in Chapter Energy Metabolism and Mitochondrial Function — the possible mechanistic implications are formalised in Hypothesis Genetic Mitophagy Vulnerability: The Accumulation Threshold Model below.

The Maccallini 2026 meta-GWAS, combining DecodeME with the Million Veteran Program (19,470 cases, 699,111 controls), has confirmed and extended these brain enrichment findings (Maccallini 2026). Beyond single-variant associations, Sardell et al. (2026) employed the PrecisionLife combinatorial analytics platform to identify synergistic SNP-SNP interactions in ME/CFS — epistatic effects invisible to standard GWAS — suggesting that gene-gene interactions contribute additional liability beyond the additive effects captured by GWAS arrays (Sardell et al. 2026). The meta-analysis replicated brain tissue enrichment across 14 brain regions + pituitary with no peripheral tissue reaching significance, identified glutamatergic synapses as the most specific replicated gene-set, and provided the first regional neuronal enrichment: independent replicated signals in distinct neuronal populations of subcortical and cerebellar regions, with a secondary signal in dopaminergic midbrain (VTA/substantia nigra).

GWAS also enables Mendelian randomisation analyses testing causal relationships between exposures and ME/CFS. Using genetic variants as instrumental variables, researchers can test whether genetically predicted inflammatory markers, vitamin D levels, or other biomarkers causally contribute to disease liability, distinguishing causation from the reverse causation and confounding that plague observational ME/CFS studies.

Hypothesis: Genetic Mitophagy Vulnerability: The Accumulation Threshold Model

Certainty: 0.35. The DecodeME identification of mitophagy (FBXL4) and ER-phagy (CCPG1) genes at genome-wide significant or near-significant loci, combined with the well-documented mitochondrial dysfunction in ME/CFS (Chapter Energy Metabolism and Mitochondrial Function), suggests a two-hit model of bioenergetic collapse.

Individuals carrying risk variants in autophagy genes have constitutionally reduced capacity to clear damaged mitochondria. Under normal conditions, this impairment is subclinical—the rate of mitochondrial damage does not exceed the reduced clearance capacity. However, an acute metabolic stress (viral infection, surgery, prolonged physical/psychological stress) produces a burst of mitochondrial damage that overwhelms the already-reduced clearance system. Damaged mitochondria accumulate past a critical threshold, producing a self-reinforcing cycle: dysfunctional mitochondria generate excess reactive oxygen species, which damage neighbouring mitochondria, further increasing the clearance burden.

If this hypothesis is correct, then: (1) ME/CFS patients should show elevated markers of impaired mitophagy (accumulation of mitochondrial DNA lesions, reduced PINK1/Parkin pathway activity, increased mitochondrial fragmentation) compared to healthy controls matched for FBXL4 genotype; (2) the severity of mitochondrial dysfunction should correlate with FBXL4 and CCPG1 genotype in a dose-dependent manner; (3) interventions that enhance mitophagy (e.g., urolithin A, NAD+ precursors, exercise-mimetic compounds that activate AMPK without requiring physical exertion) should preferentially benefit patients with autophagy risk variants; (4) the bioenergetic crisis should be partially reversible if the accumulated damaged mitochondria can be cleared, predicting that mitophagy-enhancing therapies could restore function even in established disease.

Treatment implications: Compounds proposed to enhance mitophagy (urolithin A, spermidine, NAD+ precursors such as nicotinamide riboside) could be stratified by FBXL4/CCPG1 genotype, enabling a pharmacogenomic approach to ME/CFS treatment. This represents a testable bridge between the genetic findings of DecodeME and the metabolic dysfunction documented by exercise physiology studies.

Limitations: The FBXL4 locus is sub-threshold in the main DecodeME analysis and the gene assignment is not definitive. No study has directly measured mitophagy capacity in ME/CFS patients. The two-hit threshold model is conceptually appealing but has not been tested in any chronic fatigue condition. The proposed interventions (urolithin A, NAD+ precursors) have not been trialled specifically in ME/CFS with genotype stratification.

4 Maccallini 2026 Meta-GWAS: Brain-First Genetic Architecture

Achievement: Maccallini Meta-GWAS: Brain-First Genetic Architecture

Certainty: 0.50 (preprint; not yet peer reviewed). The Maccallini 2026 meta-analysis combines DecodeME with the Million Veteran Program (MVP) GWAS, yielding a total of 19,470 ME/CFS cases and 699,111 controls—the largest genetic investigation of ME/CFS to date (Maccallini 2026). The study provides four converging lines of evidence supporting a brain-first genetic architecture.

First, tissue enrichment across 30 tissue types found significant enrichment exclusively in 14 brain regions and pituitary, with no peripheral tissue reaching significance. This replicates and extends the DecodeME brain enrichment finding (Achievement DecodeME Brain Tissue Enrichment) and argues against a primary peripheral genetic origin. Second, gene-set analysis replicated neuronal and synaptic gene sets, with glutamatergic synapses emerging as the most specific replicated signal—converging with the DecodeME identification of SHISA6 and UNC13C (Section Replication Status: Not Yet Replicated (By Design)). Third, cell-type enrichment identified independent replicated signals in distinct neuronal populations of subcortical and cerebellar regions, providing the first regional neuronal enrichment of ME/CFS genetic architecture. Fourth, a secondary signal in dopaminergic midbrain (ventral tegmental area/substantia nigra) was identified, with convergent support from both common (GWAS) and rare (WGS) variant data—rare variant analysis independently implicating synaptic genes (Birch et al. 2025).

The absence of any peripheral tissue enrichment has broad implications: if ME/CFS were primarily a peripheral immune or metabolic disorder, genetic risk variants should be detectable in those tissues. Their absence points to a brain-first model where primary genetic liability operates through neuronal circuits, and peripheral pathology arises as a downstream consequence.

Meta-analysis of DecodeME + MVP: n=19,470 cases, n=699,111 controls. Tissue enrichment in 30 tissues, cell-type analysis, gene-set replication.

Warning: Single-Anchoring Risk and Methodological Limitations

The meta-analysis is a preprint (not peer-reviewed)
The meta-analysis partially overlaps with DecodeME samples (not fully independent replication)
The MVP cohort is ~90% male vs ME/CFS patient population ~75% female, potentially biasing results, masking female-specific risk variants, and producing tissue enrichment patterns that differ from a sex-matched cohort
Organ-based enrichment methods have known annotation biases (GTEx brain samples have more sub-regions than peripheral tissues)

Limitation: SynGO Enrichment Is Unpublished, Method-Dependent, and Vulnerable to Gene-Level Confounds

The postsynaptic-assembly localization of the ME/CFS genetic signal rests on a single, unpublished MAGMA gene-set enrichment run against the SynGO resource, performed by one investigator and documented on a science blog rather than in a peer-reviewed publication, with no deposited analysis code or exact parameter set — it is not independently reproducible from its own source and has not been independently replicated. Gene-set enrichment is sensitive to analytic choices (SNP-to-gene window size, background gene set, multiple-testing method), and the DecodeME+MVP “concordance” is not fully independent, since the meta-analysis incorporates DecodeME (the same cohort counted in both runs). Two further confounds could produce a spurious postsynaptic signal without any ME/CFS-specific meaning: (1) gene-length/conservation bias — synaptic genes are exceptionally long, conserved, and loss-of-function-intolerant (Koopmans et al. 2019), and MAGMA enrichment is biased toward exactly these properties by statistical construction, so the signal may tag generic gene features rather than synapse biology; and (2) pleiotropy — SynGO terms are overrepresented among genes for intelligence, educational attainment, ADHD, autism, bipolar disorder, and neurodevelopmental disorders (Koopmans et al. 2019), so the enrichment may reflect a generic brain-expressed-conserved-gene tag shared with many neuropsychiatric traits rather than an ME/CFS-specific mechanism. The SynGO resource itself is an in silico gene-ontology knowledge base built by expert curation (which may over-annotate the most-studied postsynaptic and glutamatergic structures); any enrichment finding is an inference from GWAS gene-set analysis, with a translation gap between ontology annotation and disease tissue.

(Certainty: N/A — this is a methodological limitation. Severity applicability: unknown.)

Consequence: Until the SynGO enrichment is published with code, independently replicated in a non-overlapping cohort, and shown to survive correction for gene length/conservation and to exceed the generic brain-trait pleiotropy baseline, the postsynaptic-assembly localization should be treated as a hypothesis for testing — not as an established feature of ME/CFS — so that readers do not mistake a single unpublished analysis for a replicated genetic finding.

Hypothesis: Glutamatergic Synaptic Dysfunction as Genetically-Driven Core Mechanism

Certainty: 0.60. Replicated enrichment of glutamatergic synapse genes across both DecodeME (single-gene level via SHISA6, UNC13C) and the Maccallini meta-GWAS (gene-set level, Bonferroni-corrected) provides convergent evidence that glutamatergic signaling dysfunction is a genetically grounded component of ME/CFS, not an epiphenomenon (Maccallini 2026) (DecodeME Consortium, Ponting, et al. 2025). This aligns with the excitatory/inhibitory imbalance model proposed independently from immunological and neuroimaging data (Section DecodeME Brain Tissue Enrichment; Chapter Neurological and Neurocognitive Dysfunction) (Wirth and Scheibenbogen 2026).

Falsifiable predictions: (1) iPSC-derived neurons from high-risk genotype carriers should show altered glutamatergic synaptic transmission (altered mEPSC frequency/amplitude, aberrant AMPA/NMDA ratios) compared to low-risk carriers. (2) PET with glutamate receptor ligands (e.g., ^18F-FPEB for mGluR5) should reveal altered receptor density or occupancy in ME/CFS, with effect size correlating with glutamatergic PRS. (3) ME/CFS patients stratified by high glutamatergic PRS should show differential response to glutamatergic modulators (memantine, riluzole, lamotrigine) in a randomised trial, with greater benefit in the high-PRS subgroup.

Hypothesis: Two-Hit Model: Glutamatergic Vulnerability + Infectious Trigger

Certainty: 0.50. Glutamatergic synaptic genes (SHISA6, UNC13C, plus sub-threshold hits) identified by DecodeME and the Maccallini meta-GWAS establish a constitutional hyperexcitability substrate affecting cortico-cerebellar and brainstem circuits (DecodeME Consortium, Ponting, et al. 2025) (Maccallini 2026). Under normal conditions, this genetic vulnerability is subclinical — synaptic E/I balance is maintained within a tolerable range by homeostatic mechanisms. An infectious trigger (EBV, SARS-CoV-2, other) activates innate immune cascades that, via cytokine-mediated blood-brain barrier disruption and microglial activation, deliver an additional glutamatergic insult: IL-1$\beta$ and TNF-$\alpha$ increase synaptic glutamate release and impair astrocytic glutamate reuptake (Wirth and Scheibenbogen 2026). In genetically vulnerable individuals, this superimposed immune-mediated excitation pushes the system past a critical E/I threshold from which homeostasis cannot recover. The chronicity arises because the genetic substrate is fixed and the immune trigger produces sustained neuroinflammation — neither factor resolves independently once the destabilising synergy is established.

Falsifiable predictions: (1) iPSC-derived neurons from high-glutamatergic-PRS carriers should show greater excitotoxic vulnerability to inflammatory cytokines (IL-1$\beta$, TNF-$\alpha$) compared to low-PRS carriers; (2) PET imaging during acute infectious illness should reveal greater glutamate receptor availability changes in ME/CFS-susceptible individuals; (3) early post-infectious treatment with glutamatergic modulators (memantine, lamotrigine) should reduce progression to chronic ME/CFS in high-PRS individuals.

Hypothesis: Brain-First Genetic Model

Certainty: 0.55. The Maccallini meta-GWAS demonstrates that ME/CFS genetic risk variants are enriched exclusively in brain and pituitary tissues, with no peripheral tissue reaching significance (Maccallini 2026). This pattern—genetic liability operating primarily through neuronal circuits—contrasts with peripheral immune hypothesis expectations, which would predict enrichment in blood, spleen, or lymph node.

The brain-first model does not exclude peripheral pathology; it posits that genetic vulnerability originates in the CNS and produces peripheral manifestations through efferent signaling (autonomic dysregulation, neuroendocrine disruption, altered descending pain modulation). This is consistent with the absence of genetic correlation with classical autoimmune diseases (Section DecodeME Brain Tissue Enrichment) and with neurological abnormalities documented in Chapter Neurological and Neurocognitive Dysfunction.

Falsifiable prediction: Partitioned heritability analysis will show that common-variant heritability for ME/CFS is significantly higher in brain tissues than in blood, lymphoid, or other peripheral tissues after controlling for eQTL sharing. Falsified if peripheral tissue heritability is comparable to brain tissue heritability. Additionally, a tissue-specific Mendelian randomisation study showing that a peripheral immune biomarker (e.g., NK cell cytotoxicity) mediates ME/CFS risk independently of brain eQTL effects would refute the brain-first architecture.

Limitations: (1) Tissue enrichment reflects cumulative expression across all cell types and states; rare cell types or state-dependent expression may be missed. (2) GWAS captures common variants only; rare variants affecting peripheral tissues are not detected. (3) Brain enrichment does not identify which specific circuits are primarily affected—cell-type resolution and functional validation are required. (4) Cell-type enrichment analyses identify MSNs/eMSNs as the most specific hit but this resolution is method-dependent; MSN enrichment is shared across multiple brain-relevant traits (schizophrenia, depression, alcohol consumption, sleep duration — (Duncan et al. 2025)), so the brain-first model does not yet specify which neuronal population mediates ME/CFS risk. (5) The immune cell enrichment null result — zero significant associations across all immune cell types tested (T cells, B cells, NK cells, macrophages) via the Finucane 2018 stratified LDSC pipeline applied to the ImmGen reference dataset on DecodeME+MVP data — constrains but does not exclude peripheral immune involvement: immune pathology may be acquired rather than genetically encoded, or effect sizes may be below GWAS detection threshold. The null specifically means common-variant genetic risk does not enrich in immune cells; it does not mean immune dysfunction is absent from the disease.

Open Question: Cell-Type-Specific PRS for Biologically Coherent Subtyping

The Maccallini meta-GWAS provides the first regional neuronal enrichment of ME/CFS genetic architecture, with independent replicated signals in distinct neuronal populations of subcortical and cerebellar regions, a secondary signal in dopaminergic midbrain, and autophagy-related genes (Maccallini 2026). Could polygenic risk scores stratified by functional category — neuronal (glutamatergic) PRS, autophagy/mitophagy PRS, immune PRS — define biologically coherent subtypes with distinct clinical features, treatment responses, and prognoses? This requires cell-type-specific eQTL reference panels and partitioned heritability analysis to determine whether genetic architecture partitions into separable biological dimensions or represents a single continuous liability distribution.

Key test: Partitioned heritability analysis will show significant heterogeneity in PRS component loadings across individuals, with identifiable subgroups showing predominantly glutamatergic, autophagy, or immune genetic signatures.

Open Question: Synaptic-Density PET as a Presynaptic Proxy for Synaptic Structure

If ME/CFS genetic risk localizes to synaptic assembly and scaffolding, synaptic density itself may be measurably altered in the disease — a testable prediction for in-vivo synaptic-density imaging. Positron-emission tomography using radiotracers for the synaptic vesicle glycoprotein 2A (SV2A) — [11C]-UCB-J and [18F]-SynVesT-1 — is established as a research tool in neuropsychiatric disorders, where SV2A signal is reduced in schizophrenia cortex (Onwordi et al. 2020) and couples to glutamatergic markers in healthy volunteers (Onwordi et al. 2021); the method and its specificity limitations are reviewed in (Asch et al. 2024) and (Serrano et al. 2022). A critical design caveat is compartment specificity: SV2A is a presynaptic vesicle glycoprotein, whereas the SynGO genetic signal localizes to postsynaptic assembly and scaffolding (Unpublished SynGO Enrichment Raises a Provisional Postsynaptic-Assembly Hypothesis). Because pre- and postsynaptic structures co-assemble, SV2A signal may still covary with total synapse number and remain informative as a presynaptic proxy, but a null SV2A result would not falsify a postsynaptic-assembly deficit. An SV2A-PET study in ME/CFS — e.g., comparing across striatum, hypothalamus, and cortex against healthy controls — should therefore be designed compartment-aware (paired with a glutamatergic-neurochemistry readout such as MRS glutamate, which measures pooled synaptic glutamate, (Thapaliya et al. 2025)) rather than treated as a stand-alone test of the postsynaptic hypothesis.

(Certainty: N/A — this is an open research question / proposed tool, not an established finding. Translation gap: SV2A-PET evidence is from schizophrenia/healthy-volunteer and methodological cohorts, not ME/CFS. Severity applicability: unknown.)

Falsifiable prediction: An in-vivo [11C]-UCB-J or [18F]-SynVesT-1 PET study, ideally paired with a postsynaptic-sensitive readout, will show a measurable difference in SV2A binding (synaptic density) in one or more brain regions in ME/CFS versus matched controls. Falsified if synaptic density is indistinguishable from controls across all regions examined; a null SV2A result alone, however, does not rule out a postsynaptic-assembly deficit.

Consequence: A positive SV2A-PET result would give ME/CFS a direct, quantitative measure of synaptic density in the living brain — strengthening the claim that ME/CFS involves altered synaptic structure — and could serve as an objective severity or treatment-response biomarker. But because the tracer is presynaptic while the genetic signal is postsynaptic, a negative scan must not be read as ruling out the synaptic hypothesis. It is early-stage: SV2A-PET is an established method in other disorders but untested in ME/CFS.

Speculation: Serum Kynurenic Acid/Quinolinic Acid Ratio as Glutamate-Status Biomarker

The kynurenine pathway bridges immune activation and glutamatergic neurotransmission. Pro-inflammatory cytokines activate indoleamine 2,3-dioxygenase (IDO), shunting tryptophan toward kynurenine metabolism rather than serotonin synthesis. Downstream, kynurenic acid (KYNA) acts as an NMDA receptor antagonist (neuroprotective), while quinolinic acid (QUIN) is an NMDA agonist (neurotoxic). The KA/QA ratio therefore reflects net glutamatergic tone: a low ratio predicts glutamatergic hyperexcitability (Groven et al. 2021) (Kavyani et al. 2022). If ME/CFS involves glutamatergic dysregulation from genetically driven synaptic vulnerability, serum KA/QA ratio (HPLC-MS/MS) could serve as a peripheral proxy for central glutamatergic status, enabling treatment stratification for glutamatergic modulators (memantine, lamotrigine).

(Certainty: 0.40)

Falsifiable prediction: ME/CFS patients with low serum KA/QA ratio (< median of healthy distribution) will show greater clinical response to glutamatergic modulation (memantine) compared to high-KA/QA patients in a stratified randomised trial. Low KA/QA should also correlate with high glutamatergic PRS.

Speculation: VTA/SN Dopaminergic Effort-Motivation Collapse

Maccallini 2026 identified a secondary dopaminergic midbrain signal (VTA/substantia nigra) with convergent support from both common-variant GWAS and rare-variant WGS data (Maccallini 2026). Dopamine encodes effort cost — how much a given action is “worth” metabolically. If dopaminergic midbrain dysfunction tilts effort valuation toward “not worth it,” this reframes ME/CFS “fatigue” as a computational problem in effort-reward valuation rather than muscle exhaustion, consistent with the effort-cost miscalibration documented in cortico-striatal circuits (Chapter Neurological and Neurocognitive Dysfunction).

(Certainty: 0.40)

Falsifiable prediction: A dopamine-dependent effort-based decision-making task (e.g., progressive ratio task with fMRI) will show steeper effort discounting in ME/CFS vs controls, with discounting rate correlating with DA-PRS. Patients with high DA-PRS will show greater effort-cost sensitivity despite normal peripheral energy metabolism markers.

Speculation: Enteric Glutamatergic Dysfunction as Genetic Link to IBS r_g = 0.75

The strongest genetic correlation with ME/CFS is IBS at $r_g = 0.75$ (Maccallini 2026). The enteric nervous system uses glutamate as its primary excitatory neurotransmitter — the same genes, same synaptic machinery as the CNS. If glutamatergic risk variants affect enteric circuits, this would explain both the IBS genetic overlap and GI symptoms in ME/CFS without requiring a gut-to-brain or brain-to-gut directionality assumption: the same genetic liability operates in both compartments simultaneously.

(Certainty: 0.35)

Falsifiable prediction: ME/CFS+IBS patients will show significantly higher glutamatergic PRS than ME/CFS-only patients, independent of depression or anxiety comorbidity. Glutamatergic PRS will correlate with GI symptom severity (visceral hypersensitivity, dysmotility) within the ME/CFS+IBS subgroup.

Speculation: Dopaminergic Midbrain PRS for Effort-Motivation Subtyping

If VTA/SN dysfunction drives effort-motivation collapse (Speculation VTA/SN Dopaminergic Effort-Motivation Collapse), a dopamine-pathway-specific polygenic risk score (DA-PRS) could identify patients whose “fatigue” is primarily a motivational-effort valuation deficit rather than a metabolic-muscular limitation. DA-PRS would complement glutamatergic PRS (Hypothesis Glutamatergic Synaptic Dysfunction as Genetically-Driven Core Mechanism) by capturing a distinct dimension of genetic liability.

(Certainty: 0.40)

Falsifiable prediction: DA-PRS predicts effort-based decision-making task performance (willingness to exert effort for reward) better than Glu-PRS in ME/CFS, while Glu-PRS better predicts sensory hypersensitivity and PEM severity.

Speculation: Composite Neuronal Risk Score for Subtype Discrimination

A composite polygenic risk score combining Glu-PRS + DA-PRS + cerebellar PRS (Cer-PRS) from Maccallini cell-type enrichment data may provide better subtype discrimination than any single PRS alone (Maccallini 2026). This is analogous to polygenic risk scores combining multiple trait-specific scores for cardiovascular disease risk prediction.

(Certainty: 0.45)

Falsifiable prediction: Composite neuronal PRS (Glu + DA + Cer) will show significantly better discrimination between clinically defined ME/CFS subtypes (e.g., cognitively-predominant vs pain-predominant vs autonomic-predominant) compared to any single PRS, with AUC improvement > 0.10 over best single PRS.

Speculation: Combinatorial SNP Interactions as Clinical Prediction Tool

Sardell et al. (2026) employed the PrecisionLife combinatorial analytics platform to identify synergistic SNP-SNP interactions in ME/CFS, capturing epistatic effects invisible to standard additive GWAS models (Sardell et al. 2026). These combinatorial markers may outperform additive polygenic risk scores for clinical ME/CFS risk prediction.

(Certainty: 0.50)

Falsifiable prediction: A combinatorial epistatic risk score will show AUC improvement > 0.15 above the best additive PRS in predicting ME/CFS case-control status (p < 0.01), in an independent validation cohort not used for discovery.

Speculation: Unpublished SynGO Enrichment Raises a Provisional Postsynaptic-Assembly Hypothesis

An unpublished, single-investigator gene-set enrichment analysis of SynGO synapse-ontology terms against ME/CFS GWAS data (DecodeME, and the DecodeME+MVP meta-analysis) — documented on the ME/CFS Science Blog with no deposited code or parameter set — raises the provisional hypothesis that the glutamatergic genetic signal, already established at the gene-set level (Glutamatergic Synaptic Dysfunction as Genetically-Driven Core Mechanism), may lean toward synaptic assembly and scaffolding rather than neurotransmitter release. The most significant SynGO terms were “synapse adhesion between pre- and post-synapse” (GO:0099560), “regulation of postsynaptic density assembly” (GO:0099151), and “postsynaptic density membrane” (GO:0098839), reported as Bonferroni-significant in the DecodeME run; the DecodeME+MVP run was directionally concordant, though it is not an independent replication because the meta-analysis incorporates DecodeME. The enriched genes include postsynaptic density and trans-synaptic adhesion machinery (LRRC7, NLGN1, SHISA6, DCC) alongside genes that do not sort cleanly into that category (CACNA1E is a presynaptic calcium channel; NRXN1 is presynaptically anchored; ARFGEF2 is vesicle/trafficking-associated) — so the postsynaptic-vs-presynaptic distinction is a hypothesis to be tested, not an established exclusion. The SynGO resource itself documents that synaptic genes are exceptionally conserved and intolerant to loss-of-function mutation, and that SynGO terms are overrepresented among gene sets for neurodevelopmental and psychiatric disorders (Koopmans et al. 2019). This genetic (constitutional) account is complementary to — and not exclusive of — an acquired reduction in synaptic contacts via complement-mediated synaptic pruning, proposed separately as a mechanism for cognitive symptoms in ME/CFS (Excess Synaptic Pruning as a Substrate for Cognitive Symptoms): the two could coexist as a genetic substrate plus an acquired insult on the same synaptic-density axis.

(Certainty: 0.30 — this is a third-order inference (preprint meta-GWAS → unpublished sub-term enrichment → a postsynaptic-vs-presynaptic differential, the finest and most bias-prone claim); the direct enrichment result is unpublished, uncited, and vulnerable to gene-length/conservation bias and pleiotropy, so it is rated below the Maccallini preprint anchor ((Maccallini 2026)) it refines. Translation gap: SynGO is an in silico gene-ontology resource; the ME/CFS enrichment is inferred from GWAS gene-set analysis, not from ME/CFS tissue. Severity applicability: unknown — genetic findings not stratified by severity. Confound caveats: see SynGO Enrichment Is Unpublished, Method-Dependent, and Vulnerable to Gene-Level Confounds.)

Falsifiable prediction: A published, pre-registered MAGMA or comparable gene-set replication on an independent ME/CFS GWAS cohort (non-overlapping with DecodeME/MVP) will reproduce enrichment of the postsynaptic assembly/scaffolding SynGO terms (GO:0099560, GO:0099151, GO:0098839) at family-wise-significant levels, and this postsynaptic-vs-presynaptic specificity must survive correction for gene length/conservation (e.g., length-matched null gene sets or conditional analysis) and exceed a matched brain-trait pleiotropy baseline. The hypothesis is falsified if an independent replication shows no postsynaptic-assembly enrichment, if the signal concentrates in presynaptic vesicle-release terms, or if it attenuates to null once gene-length/conservation bias is removed.

Consequence: If this postsynaptic-assembly localization is confirmed by published, reproducible replication, it would narrow where in the brain’s wiring ME/CFS genetic risk operates — from “glutamatergic synapses” toward the machinery that builds and scaffolds postsynaptic connections — but it currently rests on an unpublished, unreproducible analysis with indirect support, so no mechanistic or drug-target conclusions should be drawn from it yet.

{{/* Phase 10a retroactive synthesis: brain-first glutamatergic genetic architecture (condenses ~20 environments across ch12/ch14d/ch14h/ch16) */}}

Synthesis: A Brain-First, Glutamatergic Genetic Architecture for ME/CFS

Taken together, the DecodeME single-gene results and the Maccallini 2026 meta-GWAS/WGS point toward a coherent genetic model that would reframe where ME/CFS liability originates — though the two are not independent lines: the Maccallini meta-analysis incorporates DecodeME (combining it with the Million Veteran Program), so the single-gene lookups and the meta-analytic gene-set signal are overlapping analyses of partly shared data, not two independent replications. The anchoring finding, reported by the Maccallini 2026 meta-analysis, is tissue specificity: genetic risk is enriched exclusively in brain and pituitary tissue across thirty tissue types, with no peripheral immune or metabolic tissue reaching significance (Brain-First Genetic Model, Maccallini Meta-GWAS: Brain-First Genetic Architecture) — a brain-first architecture formalized as the efferent-coupling causal model in Chapter Causal Hierarchy: Root Causes, Amplifiers, and Consequences (Glutamate-GABA E/I Balance ODE: Glutamate Trap Bistability). This enrichment is consistent with the CNS energy crisis as a trigger-capable root cause (Harmful Advice: The “Power of Positive Thinking”) but does not select it: brain-tissue enrichment localizes risk to neuronal tissue without discriminating among brain-centric causal models. The most specific pathway signal is glutamatergic — appearing in both the DecodeME single-gene lookups (SHISA6, UNC13C) and the meta-analysis’s Bonferroni-corrected gene-set (Glutamatergic Synaptic Dysfunction as Genetically-Driven Core Mechanism), though these are overlapping rather than independent analyses (see above) — motivating a two-hit model in which a constitutional excitatory/inhibitory-imbalance substrate is tipped past a homeostatic threshold by an infection-driven glutamatergic insult (Two-Hit Model: Glutamatergic Vulnerability + Infectious Trigger). If this signature holds, it could offer one contributing reading of several otherwise-disconnected phenomena — each of which is multiply determined and not resolved by genetics alone: it would help ground the cerebellar cognitive-affective (CCAS) phenotype in the cerebellar neuronal enrichment reported by the meta-analysis (Maccallini Cerebellar Signal Maps to CCAS Cognitive Phenotype); it offers one candidate account of the depression paradox — high genetic correlation but no shared causal variants — as shared glutamatergic genes expressed in different circuits (Depression Paradox: High r_g, No Shared Causal Variants, Depression: Shared Glutamatergic Genes, Divergent Circuit Expression); it offers one genetic reading of the strong IBS correlation via enteric glutamatergic signaling, alongside enteric-immune and autonomic contributions (Enteric Glutamatergic Dysfunction as Genetic Link to IBS r_g = 0.75); and it supports disciplined cross-disease comparisons to schizophrenia, epilepsy, and autism where the same pathway, different circuit logic prevents naive drug-repurposing inferences (Schizophrenia GWAS Glutamatergic Parallel — Same Pathway, Different Circuits, Epilepsy: Glutamatergic Hyperexcitability Threshold, Autism: Cerebellar Development and Glutamatergic Circuits). A secondary dopaminergic-midbrain signal additionally raises the possibility that “fatigue” involves an effort-valuation computation rather than only muscle exhaustion (VTA/SN Dopaminergic Effort-Motivation Collapse).

What the genetics does not yet establish is equally important. The meta-analysis is a preprint (certainty 0.50); heritability is modest ($h^2_{\text{SNP}} \approx 0.095$), so genetic liability is one contributor among many rather than a deterministic cause; brain-tissue enrichment localizes risk to neuronal circuits but does not identify which circuits are causal versus downstream; and every subtyping proposal built on this architecture — glutamatergic, dopaminergic, cerebellar, and composite polygenic risk scores (Composite Neuronal Risk Score for Subtype Discrimination, Dopaminergic Midbrain PRS for Effort-Motivation Subtyping, Combinatorial SNP Interactions as Clinical Prediction Tool, Cell-Type-Specific PRS for Biologically Coherent Subtyping) — remains untested against clinical outcomes. The peripheral glutamate proxy (Serum Kynurenic Acid/Quinolinic Acid Ratio as Glutamate-Status Biomarker) is unvalidated. Whether ME/CFS even has a brain-first, glutamatergic genetic component remains the primary open question: a preprint meta-analysis (incorporating DecodeME) and DecodeME’s own single-gene lookups are consistent with this architecture but await peer review and independent replication in a non-overlapping cohort. Only if that is confirmed does the next question arise — whether the component defines actionable patient subtypes and points to circuit-specific glutamatergic therapies, or is instead a broad constitutional risk factor whose expression is dominated by non-genetic triggers.

Consequence: The current genetic evidence — one preprint meta-analysis and the DecodeME data within it — is suggestive that ME/CFS risk may be wired more into brain glutamate-signaling circuits than into the immune system or muscles; if it survives peer review and replicates in an independent cohort, it would shift where researchers look for causes and drugs (toward brain excitation/inhibition balance) and could eventually let a blood-based DNA score flag people whose illness is primarily neuronal. But heritability is modest, the two supporting analyses share data, and none of the proposed genetic subtypes has been shown to predict who responds to which treatment — so this is a tentative reorientation of the research map, not yet a clinical tool.

{{/* Phase 3 synthesis: three-line genetic convergence on neuronal biology (cell-type enrichment findings, 2026-07-21) */}}

Synthesis: Three-Line Genetic Convergence on Neuronal Biology

Three independent evidence lines — using different data types and analytic approaches — converge on neuronal biology as the genetic substrate of ME/CFS: (1) common variant GWAS enriches in brain tissues and neuronal gene-sets (DecodeME (DecodeME Consortium, Ponting, et al. 2025), Maccallini 2026 meta-GWAS (Maccallini 2026)); (2) rare variant burden from whole-exome sequencing concentrates in neuronal and synaptic genes (Snyder, Zhao, et al. 2025); (3) cell-type enrichment analyses find medium spiny neuron (MSN) enrichment across three independent atlases and two methods while finding zero immune cell-type enrichment (Finucane et al. 2018).

This three-pronged genetic architecture (brain-enriched common variants + rare neuronal variants + immune cell-type null) has two separable implications. First, it provides converging support for the brain-first genetic model: genetic liability operates primarily through neuronal circuits, and peripheral pathology may arise as a downstream consequence rather than a primary genetic risk. Second, the immune null — the finding that ME/CFS-associated genetic variants show no enrichment in any immune cell type (T cells, B cells, NK cells, macrophages) when tested against the ImmGen reference dataset via the Finucane 2018 stratified LDSC pipeline — does not contradict the extensive evidence for immune dysfunction in ME/CFS documented in Chapter Immune System Dysfunction. It suggests instead that immune pathology in ME/CFS may be acquired (infection-triggered) or environmentally driven, with effect sizes below GWAS detection threshold, rather than genetically encoded in common variants.

GWAS cell-type enrichment tells us where constitutional liability is encoded (neurons), not which mechanisms are operative in the established disease. All trigger mechanisms — viral infection, GPCR autoantibody production, mast cell activation, TRPM3 channelopathy, skeletal asymmetry — and all amplifier mechanisms — the PEM metabolic danger cascade, kindling-driven threshold lowering, glymphatic failure, complement-mast cell amplification loops, endothelial senescence, NET/DNase imbalance — remain valid acquired pathophysiological processes. The brain-first genetic architecture does not displace them; it constrains the upstream origin of vulnerability while the triggers and amplifiers describe the downstream acquired pathology that propagates from or triggers upon that vulnerable CNS substrate. This is formalised in the two-hit model: genetic vulnerability in neuronal circuits (hit 1) + environmental trigger (hit 2) → disease, with amplifiers sustaining chronicity. Neither invalidates the other.

However, cell-type enrichment at the MSN/eMSN level is method-dependent (certainty ~0.40–0.50 versus 0.80 for the broad neuronal signal), and MSN enrichment is shared across multiple brain-relevant traits (schizophrenia, depression, alcohol consumption, sleep duration — (Duncan et al. 2025)). The genetic data identify where ME/CFS risk is encoded (neurons), and suggest which cell type may be most involved (MSNs), but does not yet identify which mechanism within those neurons mediates risk. The most important open question is whether the genetic signal points to a specific MSN-intrinsic pathology or to a broader neuronal process that MSN transcriptomes happen to tag by their high transcriptional diversity.

Consequence: Multiple independent methods using different data types agree that ME/CFS risk is genetically encoded in neurons, not immune cells — a finding that, if sustained by peer review and independent replication, would redirect ME/CFS research toward neuronal circuit mechanisms and away from the assumption that immune dysfunction is genetically primary. For patients: strengthens the evidence that ME/CFS has a biological, brain-based genetic architecture — the variation is in the genes that build and operate brain circuits, not in genes that influence psychology or behaviour.

5 Kerrebijn 2026 Fibromyalgia GWAS: CNS Genetic Architecture of a Key Comorbidity

The largest multi-ancestry GWAS of fibromyalgia (Kerrebijn et al. 2026, Nature Medicine, 2.56 million individuals) identifies 26 risk loci and demonstrates exclusive brain/neural-cell-type heritability enrichment, defining fibromyalgia genetically as a central-nervous-system (nociplastic) disorder (Kerrebijn et al. 2026). Because fibromyalgia co-occurs with ME/CFS in 20–70% of cases and both conditions show brain-enriched genetic architecture, this evidence bears directly on the shared genetic-vulnerability question raised by the DecodeME and Maccallini findings above (Sections Genetic Mitophagy Vulnerability: The Accumulation Threshold Model, Replication Status: Not Yet Replicated (By Design)).

Achievement: Kerrebijn 2026 Fibromyalgia GWAS: 26 Risk Loci and Brain-Enriched Heritability

Certainty: 0.68 (discounted from raw 0.85 by 0.80 population weight — fibromyalgia is a comorbid condition, not an ME/CFS cohort). Kerrebijn et al. (2026, Nature Medicine) conducted the largest multi-ancestry genome-wide association study (GWAS) meta-analysis of fibromyalgia to date: 2,563,755 individuals (54,629 cases, 2,509,126 controls) from 11 cohorts, identifying 26 genome-wide-significant risk loci (Kerrebijn et al. 2026).

The strongest association was a coding variant in HTT, the causal gene for Huntington’s disease (inframe glutamic-acid deletion, exon 58; OR 1.09). Gene prioritization implicated the HTT regulator GPR52 alongside diverse neural genes (DCC, DRD2/NCAM1, MDGA2, CELF4). Fibromyalgia heritability (observed scale, 10.4%) was exclusively enriched within brain tissues and neural cell types (dentate gyrus, enteric neurons, interneurons, cortical and striatal projection neurons), with no immune or glial enrichment. (Note: this observed-scale figure is not directly comparable to the liability-scale ME/CFS SNP-heritability of 9.5% in Section Replication Status: Not Yet Replicated (By Design).)

The study found strong, positive genetic correlation (rg > 0.7) with low back pain, post-traumatic stress disorder, and irritable bowel syndrome, as well as with musculoskeletal and joint-pain traits (myalgia, joint pain, hypermobility, cervicobrachial syndrome) relevant to the ME/CFS hypermobility comorbidity, and pervasive positive correlation with depression, somatoform/dissociative disorders, and migraine. Despite fibromyalgia’s 87.7% female preponderance, the inter-sex genetic correlation was 1.03 — near-identical architecture in males and females. Critically, there was no significant MHC signal and no immune/glial heritability enrichment, providing genetic evidence against a primarily autoimmune architecture (Kerrebijn et al. 2026).

Falsifiable prediction: An independent multi-ancestry fibromyalgia GWAS will replicate the majority of the 26 loci (particularly HTT/exon-58 and GPR52) and confirm exclusive brain/neural-cell-type heritability enrichment. Falsified if replication shows peripheral (immune or glial) enrichment of comparable or greater magnitude.

Consequence: This gives fibromyalgia a robust, brain-enriched genetic basis consistent with a central-nervous-system (nociplastic) role, which matters for ME/CFS because the two conditions frequently co-occur and may share a central rather than peripheral genetic vulnerability — informing the comorbidity relationship, with the direct genetic correlation still to be measured. (Severity applicability: unknown — fibromyalgia GWAS cohorts not stratified by ME/CFS severity.)

Hypothesis: Fibromyalgia as CNS Disorder with Shared Genetic Vulnerability with ME/CFS

Certainty: 0.55. The Kerrebijn 2026 GWAS found fibromyalgia’s genetic architecture exclusively enriched in brain tissues and neural cell types, which is consistent with a central nervous system (nociplastic) role rather than proof of a specific CNS mechanism (Kerrebijn et al. 2026). This converges with the ME/CFS genetic evidence from DecodeME and the Maccallini meta-GWAS, which likewise find exclusive brain/neural-cell-type heritability enrichment in ME/CFS (DecodeME Consortium, Ponting, et al. 2025) (Maccallini 2026). Both conditions therefore show a centrally-enriched genetic architecture.

ME/CFS relevance. Fibromyalgia co-occurs with ME/CFS, and both share a central-sensitization / nociplastic character (Section Cramp Origin Modestly Favours a Muscular Mechanism, with an Unproven Possible Neurogenic Contribution). The convergent brain-enriched architecture is consistent with the comorbidity reflecting a partially shared central genetic vulnerability, but the co-occurrence figure is inflated by overlapping diagnostic criteria (Wolfe 2016 absorbs ME/CFS features into fibromyalgia), so it cannot by itself establish shared biology. The confirmed shared risk regions are OLFM4 and RABGAP1L (with DCC in the fibromyalgia GWAS also reported in a prior ME/CFS overlap), which the same genes may contribute through different regulatory mechanisms (ME/CFS Science 2025b) (Kerrebijn et al. 2026). GPR52 (an HTT regulator) and HTT are fibromyalgia-risk genes whose presence in ME/CFS genetics is untested, not established.

Translation note: This is a cross-disease (fibromyalgia → ME/CFS) inference, not a cross-species one — both are human GWAS. Direct transference of specific fibromyalgia loci (HTT, GPR52, DRD2) to ME/CFS pathophysiology remains speculative; the shared-vulnerability claim rests on convergent brain-enriched architecture and confirmed locus overlap, not yet on a direct fibromyalgia × ME/CFS genetic-correlation estimate (Open Question Direct Fibromyalgia × ME/CFS Genetic Correlation Remains Untested), so the certainty reflects the indirect strength of the evidence rather than a decisive shared-genetic test.

Falsifiable predictions: (1) A cross-trait GWAS/PRS analysis between fibromyalgia and ME/CFS will show significant genetic correlation (rg substantially above zero) and shared brain-enriched heritability. (2) ME/CFS patients with comorbid fibromyalgia will carry higher fibromyalgia PRS than ME/CFS patients without it. (3) Whether HTT/GPR52 loci appear at all in ME/CFS GWAS is testable; if absent, the shared-vulnerability claim is restricted to the confirmed loci. Falsified if fibromyalgia and ME/CFS show near-zero genetic correlation or if their enrichment patterns diverge.

Consequence: If fibromyalgia and ME/CFS share a central genetic vulnerability, then treatments and biomarkers targeting central sensitization may generalize across the two conditions, and the comorbidity may be a shared-mechanism signal rather than chance co-occurrence — but this remains conditional on the untested direct genetic-correlation estimate. (Severity applicability: unknown — fibromyalgia GWAS cohorts not stratified by severity.)

Hypothesis: High Fibromyalgia Genetic Correlation With Chronic Pain and Psychiatric Traits

Certainty: 0.42. The Kerrebijn 2026 GWAS found fibromyalgia shows genetic correlation (rg) above 0.7 with low back pain, post-traumatic stress disorder, and irritable bowel syndrome, and pervasive positive correlation with depression, somatoform/dissociative disorders, and migraine (Kerrebijn et al. 2026). Independent genomic structural-equation modeling (Johnston 2025) reports nociplastic pain as a shared heritable factor underlying chronic overlapping pain conditions (Johnston, Signer, and Huckins 2025), and Lin 2026 (GenomicSEM) reports a shared latent genetic liability across fibromyalgia and psychiatric traits with synaptic-function enrichment (Lin et al. 2026).

Interpretation. These rg values establish that fibromyalgia shares substantial genetic variance with chronic pain and psychiatric/somatic traits. Whether a single transdiagnostic CNS factor drives this (versus several overlapping pathways, or shared population structure) is not yet established — the GenomicSEM common-factor analyses are a candidate formalization of that idea, not proof of it. The ME/CFS relevance is that ME/CFS already shows high genetic correlation with irritable bowel syndrome (rg = 0.75) and depression (rg = 0.60) in DecodeME (DecodeME Consortium, Ponting, et al. 2025) — the same axes implicated in the fibromyalgia findings.

Evidence-quality note: Genetic correlation is an association between heritable components, not proof of shared causal biology — correlated traits may share pathways or share population structure. Causality (a shared CNS factor driving multiple conditions) rests on the GenomicSEM/MR analyses (quasi-causal, contingent on instrument validity) rather than the rg values alone (Lin et al. 2026) (Hu et al. 2025).

Falsifiable predictions: (1) A transdiagnostic genomic SEM factor across fibromyalgia, IBS, low back pain, and depression will show significant heritability and enrichment in shared neural cell types. (2) Polygenic risk for nociplastic pain will predict symptom severity across ME/CFS, fibromyalgia, and IBS cohorts. Falsified if the shared genetic factor explains negligible variance, if enrichment is condition-specific rather than shared, or if the correlations are attributable to population structure.

Consequence: Fibromyalgia, ME/CFS, IBS, and chronic pain show high pairwise genetic correlation — a real finding that argues for shared genetic underpinnings — but whether a single central-sensitization mechanism drives this is unproven; transdiagnostic treatment strategies remain a hypothesis to test, not an established implication. (Severity applicability: unknown — GWAS cohorts not stratified by severity.)

Open Question: Direct Fibromyalgia × ME/CFS Genetic Correlation Remains Untested

The Kerrebijn 2026 fibromyalgia GWAS computed genetic correlations with chronic pain, psychiatric, and somatic disorders (rg > 0.7 with low back pain, PTSD, IBS) but did not report a direct fibromyalgia × ME/CFS genetic correlation (Kerrebijn et al. 2026). Conversely, the DecodeME ME/CFS GWAS computed rg with 3,167 traits but did not include a fibromyalgia case/control GWAS among them (DecodeME Consortium, Ponting, et al. 2025). The shared-loci inference (OLFM4, RABGAP1L, and DCC in a prior ME/CFS overlap; GPR52 is the second-nearest gene to the RABGAP1L locus, not an independent ME/CFS finding) and convergent brain-enriched architecture suggest overlap, but no single study has directly measured the genetic correlation between fibromyalgia and ME/CFS.

Why this matters. A direct rg estimate would quantify how much of the fibromyalgia–ME/CFS comorbidity is genetically shared versus coincidence or diagnostic overlap. If rg is high (as with IBS at 0.75), the two conditions are likely members of a shared chronic-overlapping-pain / neuroimmune genetic family; if rg is low despite clinical overlap, the co-occurrence may reflect convergent downstream mechanisms rather than shared germline architecture.

Falsifiable prediction: A cross-trait LD score regression between the Kerrebijn fibromyalgia GWAS and DecodeME would estimate rg; the hypothesis (shared central genetic vulnerability) predicts rg significantly above zero (point estimate > 0.3) with brain-enriched shared heritability. Falsified if rg is indistinguishable from zero.

Consequence: This is an immediately computable analysis using existing public GWAS summary statistics — it would directly answer whether ME/CFS and fibromyalgia share a genetic basis, with no new genotyping required. (Severity applicability: unknown — GWAS cohorts not stratified by ME/CFS severity.)

Hypothesis: Fibromyalgia Autoimmunity Is Not Contradicted by the Absence of an HLA Signal

Certainty: 0.30. The Kerrebijn 2026 GWAS found no significant MHC signal and no immune/glial heritability enrichment in fibromyalgia, providing genetic evidence against a primarily autoimmune (HLA-linked) architecture (Kerrebijn et al. 2026). Independent passive-transfer evidence demonstrates that IgG from fibromyalgia and long-COVID patients can reproduce symptomatology in mice, consistent with pathogenic autoantibodies (Origin: brainstorm; corpus: passive-IgG-transfer evidence in ch08 (Goebel et al. 2021)).

Reconciliation. These are not contradictory, because germline genetic architecture (HLA/MHC) and acquired humoral autoimmunity (antibodies arising post-infection) are independent layers. An absence of HLA-linked germline risk does not rule out acquired pathogenic antibodies; conversely, the passive-IgG finding does not imply an HLA-linked architecture. This mirrors the ME/CFS situation, where DecodeME found no autoimmune genetic correlation yet passive-IgG and GPCR-autoantibody evidence exists. The possibility that a fibromyalgia antibody subset maps onto the non-HLA-linked central-sensitization architecture is a framing device — a coherent way to hold both findings — not an independently tested hypothesis.

Falsifiable predictions: (1) Fibromyalgia patients who respond to immunomodulation (IVIG/plasmapheresis) will cluster in an autoantibody-positive subset that shows no enrichment for the 26 GWAS loci. (2) Passive-transfer-positive fibromyalgia sera will correlate with autoantibody titers rather than with GWAS risk-score. Falsified if the GWAS loci and the autoantibody subset are the same patients (fully overlapping), which would argue for a unified HLA-linked mechanism.

Consequence: The no-MHC GWAS finding and the passive-IgG evidence can both be true because germline genetics and acquired antibodies are separate layers — so fibromyalgia need not be forced into an “is/isn’t autoimmune” binary, and whether an antibody-defined subset exists and responds to immunomodulation remains to be tested.

Speculation: GPR52/HTT Neural Pathway as a Fibromyalgia Genetic Lead With Unproven ME/CFS Relevance

Certainty: 0.25. The Kerrebijn 2026 fibromyalgia GWAS’s strongest signal was a coding variant in HTT, with the HTT regulator GPR52 prioritized as a causal gene (Kerrebijn et al. 2026). GPR52 is an orphan G-protein-coupled receptor that negatively regulates huntingtin levels, and it is an investigational drug target in Huntington’s disease (GPR52 antagonists reduce mutant huntingtin). This is a fibromyalgia genetic lead. Whether GPR52/HTT are present in ME/CFS genetics is untested — the confirmed ME/CFS-overlap loci are OLFM4 and RABGAP1L (with DCC), not GPR52 or HTT (Origin: brainstorm). A shared neural pathway through HTT/GPR52/DRD2 is therefore a candidate mechanism contingent on ME/CFS data that do not yet exist.

Druggability caveat. This is a research/repositioning direction, NOT a treatment recommendation. GPR52-targeting compounds are largely investigational with no human safety data in ME/CFS or fibromyalgia, and the fibromyalgia HTT/GPR52 signal may be condition-specific (see Limitation Kerrebijn 2026 Fibromyalgia GWAS: Methodological Limitations). No GPR52 agent is approved or appropriate for off-label use.

Falsifiable predictions: (1) The prerequisite test: whether the HTT/GPR52/DRD2 loci appear at all in ME/CFS GWAS. If absent, the shared-neural-pathway hypothesis is falsified and GPR52/HTT are fibromyalgia-specific. (2) If present and shared, ME/CFS and fibromyalgia patients with high genetic loading at these loci should show convergent neural phenotypes (e.g., altered striatal/cortical signaling on imaging) tracking PRS. Falsified if the loci are not shared or show no functional neural correlate in ME/CFS.

Consequence: GPR52/HTT is a genetically-supported molecular lead in fibromyalgia that could eventually be druggable, but its relevance to ME/CFS is unproven and no GPR52 agent is appropriate today — a research direction, not a treatment.

Open Question: Null Reading: Fibromyalgia-ME/CFS Comorbidity as Referral/Clinical Artifact

Certainty: 0.38. The 20–70% fibromyalgia-ME/CFS co-occurrence may partly reflect referral and diagnostic artifact rather than shared biology. The 2016 Wolfe fibromyalgia criteria incorporate fatigue, unrefreshed sleep, and cognitive dysfunction, absorbing core ME/CFS features into the FM definition and inflating overlap; community-based estimates place true co-occurrence closer to 22–35% (Origin: brainstorm). If the direct fibromyalgia × ME/CFS genetic correlation (currently untested — see Open Question Direct Fibromyalgia × ME/CFS Genetic Correlation Remains Untested) turns out near zero, the comorbidity would be best explained by shared symptom-referral patterns and overlapping diagnostic criteria rather than shared germline architecture.

Falsifiable prediction: If the FM×ME/CFS genetic correlation (rg) is near zero AND the clinical overlap is concentrated in patients diagnosed via criteria that embed fatigue/cognition (Wolfe 2016), the comorbidity is substantially a criteria/referral artifact. Falsified (shared-biology supported) if rg is significantly positive and the overlap persists under strict, non-overlapping diagnostic criteria.

Consequence: This is the honest null hypothesis — it prevents over-interpreting the comorbidity as proven shared biology before the decisive genetic-correlation test is run.

Limitation: Kerrebijn 2026 Fibromyalgia GWAS: Methodological Limitations

Several limitations constrain how far the Kerrebijn 2026 fibromyalgia GWAS findings can be generalized to ME/CFS:

Ancestry bias. The meta-analysis is ~90% European; generalization and polygenic-risk-score transferability to non-European ancestries is limited (Kerrebijn et al. 2026).
Case definition. Fibromyalgia was defined by ICD-10 code M79.7 in electronic records, which may include or exclude individuals not meeting full syndromic (ACR) criteria; a “healthy participant bias” in biobank cohorts further limits representativeness.
Not ME/CFS-specific. The cohort is fibromyalgia, not ME/CFS. Extrapolation of specific loci (HTT, DRD2) to ME/CFS is indirect.
Single published meta-analysis. Not yet fully replicated locus-by-locus (partial overlap with the independent Bright 2026 GWAS (Bright et al. 2026)).
Cell-type enrichment resolution. Cell-type enrichment is limited by overlapping gene expression across related cell types, so “neural” enrichment does not cleanly isolate a single neuron population.
Autoimmune interpretation. The absence of an MHC signal argues against a primarily autoimmune genetic architecture, but genetic architecture (germline) and acquired serological autoantibody phenomena (as in the passive-IgG-transfer evidence) are different layers — the genetic finding does not refute acquired autoimmunity.

Consequence: The fibromyalgia genetic findings are robust for fibromyalgia itself but must be extrapolated to ME/CFS with explicit caveats about ancestry, case definition, and the cross-disease gap.

Limitation: GPR52 Agonist Availability and Safety Gap Blocks Clinical Translation

The GPR52 finding in the Kerrebijn 2026 fibromyalgia GWAS identifies a potential molecular target (Kerrebijn et al. 2026). However, the translational path is blocked on two grounds: (1) GPR52-targeting compounds are largely investigational (developed as Huntington’s-disease candidates), with no approved agent and no human safety or efficacy data in fibromyalgia or ME/CFS; and (2) the fibromyalgia HTT/GPR52 signal may be condition-specific — whether it appears in ME/CFS genetics at all is untested, and the confirmed ME/CFS-overlap loci are OLFM4 and RABGAP1L, not GPR52/HTT (see Limitation Kerrebijn 2026 Fibromyalgia GWAS: Methodological Limitations). No GPR52 agonist or antagonist is appropriate for clinical use in ME/CFS or fibromyalgia today; any such discussion is research-only. (Origin: brainstorm, critical category 12.)

Consequence: A genetically-supported target with no clinical path is a research direction, not a treatment — clinicians should not expect any GPR52-based therapy for the foreseeable future.

{{/* Phase 10a synthesis: shared central genetic architecture across fibromyalgia and ME/CFS */}}

Synthesis: Convergent Brain-Enriched Genetic Architecture in Fibromyalgia and ME/CFS

The fibromyalgia GWAS of Kerrebijn et al. (2026) and the ME/CFS genetic findings of DecodeME and the Maccallini meta-GWAS both find heritability enriched in brain tissues and neural cell types: fibromyalgia’s is exclusively brain/neural-enriched (Achievement Kerrebijn 2026 Fibromyalgia GWAS: 26 Risk Loci and Brain-Enriched Heritability), mirroring the brain-first genetic architecture established for ME/CFS. This convergent enrichment, together with confirmed shared risk regions (OLFM4, RABGAP1L, and DCC in a prior ME/CFS overlap), is consistent with a partially shared central genetic vulnerability (Hypothesis Fibromyalgia as CNS Disorder with Shared Genetic Vulnerability with ME/CFS). Caution is required on two fronts: the fibromyalgia-ME/CFS co-occurrence figure is inflated by overlapping diagnostic criteria (Wolfe 2016) and cannot itself establish shared biology, and GPR52/HTT are fibromyalgia-risk genes whose presence in ME/CFS genetics is untested.

A second convergent line points transdiagnostically: fibromyalgia’s genetic correlation (rg > 0.7) with IBS, low back pain, and PTSD, and pervasive correlation with depression and migraine, alongside ME/CFS’s rg = 0.75 with IBS, indicate shared chronic-overlapping-pain and gut-brain genetic variance (Hypothesis High Fibromyalgia Genetic Correlation With Chronic Pain and Psychiatric Traits). The most important open question is whether a direct fibromyalgia × ME/CFS genetic correlation is significantly positive (Open Question Direct Fibromyalgia × ME/CFS Genetic Correlation Remains Untested) — currently untested — and whether the comorbidity reflects shared biology or partly diagnostic/referral artifact (Open Question Null Reading: Fibromyalgia-ME/CFS Comorbidity as Referral/Clinical Artifact). Until that test is run, the two readings are near-parity.

What the evidence supports: both conditions show a brain/neural-enriched genetic architecture and share several risk regions. What remains speculative: whether a single central/nociplastic mechanism drives both (genetic correlation is association, not proof of shared causal biology), whether specific loci (HTT/GPR52/DRD2) transfer mechanistically to ME/CFS, and whether the no-MHC finding and the passive-IgG autoantibody evidence require reconciliation beyond “germline genetics and acquired antibodies are separate layers” (Hypothesis Fibromyalgia Autoimmunity Is Not Contradicted by the Absence of an HLA Signal).

Consequence: The convergent brain-enriched genetic architecture supports the case that ME/CFS and fibromyalgia may share a common central-nervous-system vulnerability — and makes a direct fibromyalgia × ME/CFS genetic-correlation analysis the key next test — but the shared-architecture conclusion is not yet proven, and no change to patient management follows from the genetics alone.

6 PrecisionLife Combinatorial Genetics and GLP-1 Pathway Enrichment

Limitation: Single-Source Caveat — All GLP-1 Genetic Findings From One Conference Presentation

All findings in this section derive from a single conference presentation (Gardner 2026) — not a peer-reviewed publication — by a company (PrecisionLife) with commercial interest in its combinatorial analytics platform. The genetic pathway enrichment results have not been independently replicated. The combinatorial method is proprietary and cannot be audited by independent researchers. No negative controls have been published testing whether other drug classes (e.g., SSRIs, antivirals, corticosteroids) would show comparable enrichment in the same gene set. These findings establish the genetic rationale as a hypothesis (see Hypothesis:, certainty 0.40), not as established evidence. All downstream content across chapters Immune System Dysfunction, Neurological and Neurocognitive Dysfunction, and Medications Targeting Underlying Mechanisms rests on this single unreviewed source and is qualified by explicit low-to-moderate certainty ratings.

Beyond the additive SNP associations captured by GWAS, the PrecisionLife combinatorial analytics platform has identified over 250 core ME/CFS-associated genes through synergistic SNP-SNP interactions across the DecodeME and UK Biobank cohorts (Gardner 2026) (Sardell et al. 2026) (Das et al. 2022). This platform captures epistatic effects invisible to standard GWAS methods, providing a complementary view of ME/CFS genetic architecture, though epistatic GWAS methods have a history of difficult replication in complex disease genetics due to the combinatorial explosion of tests and consequent multiple-comparison burden. The core gene set identifies multiple biological pathways with potential therapeutic relevance.

The pathways targeted by glucagon-like peptide 1 (GLP-1) receptor agonists (GLP-1 RAs) are enriched among ME/CFS-associated genes identified by this platform. These include synaptic and calcium signalling, glucose homeostasis, and endothelial dysfunction pathways — all independently implicated in ME/CFS pathophysiology (Chapters Neurological and Neurocognitive Dysfunction, Energy Metabolism and Mitochondrial Function, Cardiovascular Dysfunction). This pathway-level enrichment provides a genetic rationale for investigating GLP-1 RAs as candidate repurposing agents in ME/CFS, though no clinical data exist in this population (Gardner 2026).

GLP-1 RA-targeted pathways enriched among ME/CFS-associated genes identified by PrecisionLife combinatorial analytics (Gardner 2026).
Enriched pathway	Relevance to ME/CFS pathophysiological system (Chapter reference)
Synaptic signalling	Neuronal dysfunction, cognitive impairment, sensory hypersensitivity (Neurological and Neurocognitive Dysfunction)
Calcium signalling	NK cell TRPM3 dysfunction, impaired Ca²⁺ flux, ion channelopathy (Immune System Dysfunction)
Glucose homeostasis	Metabolic dysfunction, reduced ATP synthesis, glycolytic shift (Energy Metabolism and Mitochondrial Function)
Endothelial dysfunction	Orthostatic intolerance, cerebral hypoperfusion, microvascular pathology (Cardiovascular Dysfunction)

A further independent line of evidence comes from protective gene analysis. PrecisionLife identified alleles that are specifically under-represented in ME/CFS patients (i.e., protective). Several of these protective alleles map to type 2 diabetes, insulin-related signalling, and BMI-associated pathways (Gardner 2026). This convergence — risk alleles and protective alleles in overlapping metabolic pathways — is consistent with the hypothesis that metabolic dysregulation has a genetic component in ME/CFS. The protective allele overlap also mirrors epidemiological observations that conditions treated with GLP-1 RAs (T2D, obesity) show apparent inverse associations with ME/CFS risk in some analyses, though causal direction remains unestablished.

Hypothesis:

GLP-1 RA Pathway Enrichment as Genetic Basis for Targeted Repurposing

Certainty: 0.40. (0.35→0.40: convergence with state-dependent endothelial dysfunction literature — genetics and physiological assessment point to same endothelial vulnerability endpoint via complementary domains. All evidence extrapolated from non-ME/CFS populations; single unreviewed conference presentation source; proprietary platform; zero clinical data.)

PrecisionLife’s combinatorial genetic analysis demonstrates that the biological pathways targeted by GLP-1 receptor agonists — synaptic and calcium signalling, glucose homeostasis, endothelial dysfunction — are specifically enriched among the over 250 core ME/CFS-associated genes (Gardner 2026) (Sardell et al. 2026). The convergence of risk alleles and protective alleles (the latter overlapping with T2D/insulin-signalling/BMI pathways) provides orthogonal genetic evidence that metabolic and synaptic pathways are not merely downstream consequences of illness but may represent genetically determined therapeutic targets.

Testable falsification points: (1) Independent replication of GLP-1 pathway enrichment using non-PrecisionLife methods and independent ME/CFS genetic datasets. (2) A genetically-stratified trial of a GLP-1 RA in ME/CFS shows significant subgroup-by-treatment interaction (responders vs non-responders). Falsified if replication fails, or if GLP-1 RA shows no efficacy in any genetically-defined subgroup.

Status: Not yet replicated independently; combinatorial method proprietary; no clinical trial data; pathway enrichment not validated in ME/CFS tissue.

The genetic findings also motivate a precision medicine approach to repurposing. Because different ME/CFS patients carry different combinations of risk variants, the same drug may be effective in one patient and ineffective — or even harmful — in another. PrecisionLife’s AI-driven combinatorial analytics platform is being used to identify genetic drivers of GLP-1 RA efficacy so that strong and weak responders can be stratified before trials begin (Gardner 2026). For GLP-1 RA investigation in ME/CFS, they propose focusing on three genetically-informed subgroups: (1) autoimmune/inflammatory, (2) cardiovascular, and (3) energy metabolism. Each subgroup maps to distinct pathway enrichments and would be predicted to respond through different mechanisms.

6.1 Computable Analyses Using Existing DecodeME Data

DecodeME’s combined genotype and phenotype data (n=17,000+ with GWAS array, severity classification, diagnostic criteria, infection typing, depression comorbidity, amitriptyline use, family history, and sex) enables a programme of computational analyses that require no new recruitment or genotyping. The following hypotheses and open questions are all testable using existing DecodeME data plus publicly available external GWAS summary statistics. They are grouped by analytical approach.

6.1.1 Sex-Stratified Genetic Architecture

Hypothesis: Sex-Differential Genetic Architecture in ME/CFS

Certainty: 0.55. ME/CFS affects approximately 75% females, but whether this excess reflects a quantitative threshold difference (same genetic architecture, lower threshold in females due to hormonal or X-linked modifiers) or a qualitative difference (partially distinct loci contributing in each sex) is unknown. DecodeME’s approximately 11,700 female and 3,900 male cases enable sex-stratified GWAS and cross-sex genetic correlation ($r_{g,\text{sex}}$) estimation via LD score regression.

Autoimmune diseases show both patterns: systemic lupus erythematosus has sex-differential genetic architecture, while rheumatoid arthritis does not. DecodeME’s male sample size (n$\\approx$ 3,900) is sufficient for discovery of strong-effect sex-specific loci, comparable to early psychiatric GWAS that identified sex-differential signals.

Falsifiable prediction: If $r_{g,\text{sex}} > 0.8$, sex-differential genetic architecture is falsified and the female excess arises from non-genetic modifiers (hormonal, immune, environmental). If $r_{g,\text{sex}} < 0.6$, male and female ME/CFS are partially distinct genetic entities. The prediction is that $r_{g,\text{sex}}$ will be 0.6–0.8 (partial overlap), with immune-ambiguous loci showing more sex differentiation than neuronal loci.

Open Question: X Chromosome Association Analysis in ME/CFS

The X chromosome carries the highest density of immune-related genes in the genome, and X-inactivation escape produces higher expression of certain immune genes in XX individuals. Standard GWAS analyses typically exclude the X chromosome. An X chromosome association analysis in DecodeME—with sex-appropriate genotype coding (0/1 for males, 0/1/2 for females with dosage compensation correction)—could test whether X-linked immune gene variants (TLR7, TLR8, FOXP3, CD40LG, CXCR3) contribute to ME/CFS risk and directly explain part of the female excess. If X-linked associations are absent, the female excess must arise from autosomal loci interacting with sex hormones or other modifiers.

6.1.2 Severity as Genetic Phenotype

Hypothesis: PRS Dose-Response Across ME/CFS Severity Grades

Certainty: 0.60. If ME/CFS severity is partly genetically determined (higher genetic loading $=$ more severe disease), polygenic risk score should increase monotonically across severity grades (mild $<$ moderate $<$ severe $<$ very severe). This pattern is documented in schizophrenia (PRS predicts psychosis severity) and inflammatory bowel disease (PRS predicts need for surgery) (DecodeME Consortium, Ponting, et al. 2025). DecodeME’s severity classification across 17,000+ participants enables direct testing via ordinal logistic regression of PRS against severity, adjusted for sex, age at onset, illness duration, and infection trigger.

Additionally, Sardell cluster-specific PRS (neuronal, immune, autophagy) can be tested for differential severity associations: does one genetic pathway predict severity more strongly than others?

Falsifiable prediction: If PRS does not differ significantly across severity grades (ANOVA $p > 0.05$, variance explained $< 0.5%$), genetic loading does not determine severity and environmental factors dominate. The prediction is that neuronal-synapse cluster PRS will show the strongest severity gradient (reflecting brain circuit vulnerability as a severity determinant), while immune cluster PRS will be flat across severity grades (reflecting immune variants as susceptibility switches rather than severity modulators).

Open Question: Severe/Very-Severe Subgroup GWAS

Most ME/CFS GWAS treat all cases identically. A within-case GWAS comparing severe/very-severe cases (estimated n$\\approx$ 2,500) versus mild/moderate cases (n$\\approx$ 12,500) would remove case-control confounding entirely—both groups have ME/CFS, differing only in severity. If severity-specific loci enrich for autophagy/mitophagy pathways (the FBXL4 cluster), this would support the hypothesis that mitochondrial quality control capacity determines the severity ceiling (Hypothesis Genetic Mitophagy Vulnerability: The Accumulation Threshold Model). If they enrich for neuronal/synaptic genes, severity is determined by brain circuit vulnerability.

6.1.3 Genotype × Infection Trigger Interactions

Hypothesis: Genotype-Trigger Interaction: SNP Effects That Differ by Infection Type

Certainty: 0.45. The two-hit model (Hypothesis Two-Hit Model: Glutamatergic Vulnerability + Infectious Trigger) predicts that some genetic variants matter more for certain triggers. Using Bretherick 2023 infection typing (EBV/IM, non-EBV respiratory, Lyme/Q-fever, non-infectious), a case-only gene-environment interaction analysis at the eight genome-wide significant loci can test whether effect sizes differ across trigger categories (Cochran $Q$ heterogeneity test). The case-only design—comparing genotype frequencies across trigger subgroups within cases—is more powerful than case-control GxE and requires no new controls (DecodeME Consortium, Ponting, et al. 2025).

Falsifiable prediction: If no SNP shows significant genotype-trigger interaction (interaction $p > 0.006$ for the eight GWS loci after Bonferroni), the same genetic architecture produces ME/CFS regardless of trigger type, and the two-hit model’s prediction of trigger-specific genetic modulation is falsified. The prediction is that at least two of eight GWS loci will show significant effect-size heterogeneity across trigger groups, with immune-annotated loci (BTN2A2, RABGAP1L) showing larger effects in EBV-triggered cases.

Hypothesis: Non-Infectious Onset as Higher Genetic Loading Subtype

Certainty: 0.50. Non-infectious-onset ME/CFS ($\\approx$ 30–40% of cases) lacks the canonical “second hit.” If these patients develop ME/CFS without an identifiable environmental trigger, they may require higher genetic loading—more risk alleles at more loci—to cross the disease threshold. Alternatively, their triggers are unmeasured (subclinical infections, gradual stress accumulation). DecodeME can test this directly: if mean PRS is significantly higher in non-infectious-onset cases compared to infection-triggered cases, it supports a spectrum model where genetic risk substitutes for environmental trigger.

Falsifiable prediction: If mean PRS is not higher in non-infectious-onset cases compared to EBV-triggered cases (one-sided $t$-test $p > 0.05$), the high-genetic-loading model for non-infectious onset is falsified. The prediction is that non-infectious-onset cases will have $\\approx$ 0.15 SD higher PRS, and the genetic correlation between non-infectious and EBV-triggered ME/CFS will be 0.7–0.9 (partially overlapping but not identical architecture).

6.1.4 Comorbidity Decomposition

Hypothesis: Depression Comorbidity in ME/CFS: Genetic Predictor or Independent Phenocopy?

Certainty: 0.60. The depression paradox ($r_g = 0.60$ but no shared causal variants; Speculation Depression Paradox: High r_g, No Shared Causal Variants) can be decomposed within DecodeME by constructing a depression PRS from publicly available PGC GWAS summary statistics and applying it to DecodeME participants (DecodeME Consortium, Ponting, et al. 2025). If depression PRS predicts depression comorbidity within ME/CFS but does not predict ME/CFS core symptoms (PEM, OI, cognitive dysfunction), depression in ME/CFS patients is a genuine comorbidity with shared risk but separate pathology. If depression PRS predicts specific ME/CFS symptoms, those symptoms share biological architecture with depression. If ME/CFS PRS is identical in depressed and non-depressed ME/CFS patients, depression does not arise from the ME/CFS genetic architecture itself.

Falsifiable prediction: If depression PRS significantly predicts ME/CFS core symptoms (PEM severity, OI) independent of depression status ($\beta > 0.05$, $p < 0.01$), the “separate pathology” model is falsified and shared biology extends beyond comorbidity. The prediction is that depression PRS will predict depression comorbidity (OR $\\approx$ 1.1–1.2 per SD PRS) but not PEM or OI ($p > 0.1$), confirming that ME/CFS core features are genetically independent of depression.

Speculation: Amitriptyline Pharmacogenomics: CYP Metaboliser Status and Treatment Selection

The $r_g = 0.61$ between ME/CFS and amitriptyline use is unexplained. It could reflect prescribing indication overlap (amitriptyline targets pain, neuropathy, insomnia—all ME/CFS symptoms), pharmacogenomic selection (specific CYP genotypes tolerate amitriptyline and have ME/CFS risk), or shared biological pathways (amitriptyline’s NMDA receptor antagonism connects to the glutamatergic hypothesis). CYP2D6 and CYP2C19 metaboliser status can be imputed from DecodeME GWAS array data using Stargazer or PharmCAT pipelines. If the $r_g$ is abolished after conditioning on CYP loci, pharmacogenomic variants are the primary driver. If not, shared glutamatergic/synaptic pathway genes are more likely responsible—specifically, the same neuronal-synapse loci contributing to ME/CFS risk also increase the probability of being prescribed amitriptyline via shared pain/neuropathy phenotype.

(Certainty: 0.50)

Falsifiable prediction: If CYP2D6 metaboliser status does not associate with amitriptyline use patterns or symptom burden among users ($p > 0.05$), pharmacogenomic selection is not driving the genetic correlation.

6.1.5 Heritability Architecture

Hypothesis: Functional Annotation Heritability Partitioning: Brain-First at the Regulatory Level

Certainty: 0.65. Stratified LD score regression (S-LDSC) can partition $h^2_{\text{SNP}} = 0.095$ into functional categories: coding versus non-coding, enhancer versus promoter versus intergenic, brain-specific regulatory elements (Roadmap Epigenomics), and immune cell regulatory elements (DecodeME Consortium, Ponting, et al. 2025). If heritability concentrates in brain-specific enhancers (as in schizophrenia and bipolar disorder), the brain-first model (Hypothesis Brain-First Genetic Model) gains regulatory-genomic support. If it concentrates in immune cell enhancers, the immune model gains genetic support. If it distributes without enrichment, ME/CFS is genetically diffuse (more like height than schizophrenia). This analysis requires only summary statistics and can be completed in days using existing software.

Falsifiable prediction: If no functional category shows significant enrichment (all enrichment $p > 0.05\\/97$ after Bonferroni), ME/CFS heritability is diffusely distributed and the brain-first model lacks regulatory-genomic support. The prediction is that brain-specific enhancers (particularly cortical and cerebellar) will show $> 5 \times$ enrichment, immune enhancers will show modest enrichment ($\approx 2 \times$), and coding variants will be depleted—consistent with the regulatory architecture observed in psychiatric GWAS.

6.1.6 Mendelian Randomisation with DecodeME

Open Question: Bidirectional Mendelian Randomisation: Sleep, IBS, and Depression

DecodeME GWAS summary statistics enable two-sample bidirectional Mendelian randomisation (MR) against publicly available GWAS for sleep traits, IBS, and depression. Three analyses with distinct clinical implications:

Sleep traits: If genetic instruments for poor sleep causally increase ME/CFS risk, sleep disruption is on the causal pathway. If genetic instruments for ME/CFS causally increase sleeping too much ($r_g = 0.66$), hypersomnia is a downstream consequence.

IBS: Bidirectional MR can distinguish shared glutamatergic vulnerability (no directional causation—horizontal pleiotropy), gut-to-brain causation (IBS → ME/CFS via immune activation from gut permeability), or brain-to-gut causation (ME/CFS → IBS via vagal/autonomic disruption). If neither direction shows significance, the $r_g = 0.75$ arises from shared genetic architecture without causal mediation, supporting the enteric glutamatergic hypothesis (Speculation Enteric Glutamatergic Dysfunction as Genetic Link to IBS r_g = 0.75).

Depression: If MR shows that genetic liability to depression does not causally increase ME/CFS risk, this is strong evidence against the psychogenic model. If genetic liability to ME/CFS causally increases depression risk, depression is a downstream consequence of chronic illness (reactive depression). Latent causal variable (LCV) analysis can estimate the genetic causality proportion.

6.1.7 Genetic Correlation Decomposition

Hypothesis: Partitioned Genetic Correlation: Biological Basis of the IBS Overlap

Certainty: 0.55. The $r_g = 0.75$ with IBS is the strongest genetic correlation, but its biological basis is unknown. Local genetic correlation estimation (rho-HESS or SUPERGNOVA) across the genome can identify specific genomic regions driving the correlation and partition it into regions containing neuronal/synaptic genes, immune genes, ENS-expressed genes, or serotonergic/dopaminergic genes (DecodeME Consortium, Ponting, et al. 2025). Three competing models make distinct predictions: (a) shared glutamatergic vulnerability (enteric glutamatergic hypothesis; Speculation Enteric Glutamatergic Dysfunction as Genetic Link to IBS r_g = 0.75) predicts enrichment at glutamatergic loci; (b) shared serotonergic variants predict enrichment at HTR and TPH genes; (c) shared autonomic regulation predicts enrichment at vagal/autonomic loci.

Falsifiable prediction: If the $r_g = 0.75$ is uniformly distributed across the genome (no region contributing $> 5%$), no single pathway mediates the correlation. The prediction is that the correlation will be concentrated at $\\approx$ 20–50 genomic regions, with significant enrichment for neuronal/synaptic annotations and modest enrichment for serotonergic loci.

6.1.8 Diagnostic Criteria and Genetic Architecture

Hypothesis: CCC Versus IOM: Do Diagnostic Criteria Cut Along Genetic Boundaries?

Certainty: 0.45. DecodeME accepted patients meeting CCC (Canadian Consensus Criteria), ICC (International Consensus Criteria), or IOM criteria. CCC requires PEM plus specific neurological, autonomic, and immune criteria; IOM requires PEM but is broader. If patients meeting CCC (more restrictive) have a different genetic architecture than IOM-only patients, diagnostic criteria cut along a biological boundary. If genetic architecture is identical, the criteria differences are clinically but not biologically meaningful. Testing: GWAS separately for each diagnostic subgroup versus shared controls, followed by cross-subgroup genetic correlation estimation.

Falsifiable prediction: If the genetic correlation between CCC-only and IOM-only subgroups exceeds 0.90, the diagnostic criteria do not identify biologically distinct subgroups, and the field’s criteria debates are genetically moot. The prediction is that $r_g$ will be 0.7–0.85 (moderate divergence), with CCC-only patients showing higher neuronal-cluster PRS.

6.1.9 Cross-Trait and Cross-Disease Analyses

Open Question: Genetic Correlations with ADHD and Fibromyalgia

Two genetic correlations not yet computed from DecodeME would inform cross-disease models.

ADHD: ADHD involves dopaminergic and noradrenergic dysfunction, cognitive fatigue, and executive dysfunction—symptoms shared with ME/CFS. If $r_g$(ME/CFS, ADHD) $> 0.30$, this would support the Maccallini VTA/SN dopaminergic signal (Speculation VTA/SN Dopaminergic Effort-Motivation Collapse), suggest ADHD medications may benefit a dopaminergic ME/CFS subtype, and raise whether childhood ADHD is a risk factor for post-infectious ME/CFS.

Fibromyalgia: ME/CFS and fibromyalgia frequently co-occur. If $r_g$(ME/CFS, FM) $> 0.80$, they are genetically near-identical, suggesting a spectrum disorder. If $r_g < 0.30$, the conditions are genetically distinct despite phenotypic overlap. Partitioned $r_g$ by pain pathway genes versus neuronal/synaptic genes would reveal whether the overlap is pain-mediated or reflects shared central sensitisation.

Hypothesis: 76 Shared Long COVID Genes: Trigger-Specific or Shared Downstream Vulnerability?

Certainty: 0.50. Sardell et al. (2026) identified 76 genes shared between ME/CFS and Long COVID combinatorial analyses (Sardell et al. 2026). A “shared-gene PRS” constructed from SNPs in these 76 genes can be tested within DecodeME: does it preferentially predict post-respiratory-infection-onset ME/CFS (versus EBV-onset or non-infectious), or does it predict ME/CFS equally across trigger types? If the former, these genes represent a respiratory-post-infectious genetic subtype. If the latter, the overlap reflects shared downstream pathology (brain circuit vulnerability) rather than shared trigger biology.

Falsifiable prediction: If the shared-gene PRS does not preferentially predict post-respiratory-onset ME/CFS (interaction $p > 0.1$), the 76-gene overlap reflects shared downstream biology, not trigger-specific vulnerability. The prediction is modest trigger preference (OR $\\approx$ 1.15 for respiratory versus $\\approx$ 1.05 for EBV), but the dominant signal will be trigger-independent.

6.1.10 Between-Cluster Epistasis

Speculation: Sardell Cluster Cross-Talk: Synergy Between Neuronal and Autophagy Pathways

Sardell et al. (2026) identified within-cluster epistasis (synergistic interactions among genes in the same functional cluster) (Sardell et al. 2026). The most biologically interesting question is between-cluster interactions: does having both neuronal AND autophagy risk create a synergy beyond their additive effects? A cluster-pair interaction PRS (product of cluster-specific PRSs) can test whether the interaction term predicts ME/CFS risk or severity beyond additive effects. If the neuronal $\\times$ autophagy interaction is strongest, mitophagy failure in neurons is the critical convergence point, supporting a model where neurons with both synaptic dysfunction and impaired mitophagy are uniquely vulnerable (connecting Hypothesis Genetic Mitophagy Vulnerability: The Accumulation Threshold Model with Hypothesis Glutamatergic Synaptic Dysfunction as Genetically-Driven Core Mechanism).

(Certainty: 0.40)

Falsifiable prediction: If no between-cluster interaction term is significant (interaction $p > 0.01$ for all pairs), the clusters act independently and ME/CFS risk is additive across pathways. The prediction is that the neuronal $\\times$ autophagy interaction will show synergy coefficient $> 1.2$, while the immune cluster acts additively.

6.1.11 Family History and Genetic Subtyping Validation

Open Question: PRS Versus Family History: Measured and Unmeasured Genetic Risk

Comparing PRS between ME/CFS patients with and without a family history of ME/CFS quantifies what proportion of familial risk is captured by common variants. Family history captures all heritable factors (common variants, rare variants, structural variants, epigenetic inheritance) plus shared environment. If PRS accounts for $< 20%$ of the family history effect, rare variant studies (SequenceME WGS) are high priority. If PRS captures most familial risk, larger GWAS will be the most productive strategy. An interaction test—whether family-history-positive patients with high PRS have earlier onset and greater severity than predicted by either factor alone—probes whether genetic risk compounds non-linearly.

6.1.12 Cross-Trait PRS for Symptom Subtyping

Speculation: External PRS as Genetic Modifiers of ME/CFS Symptom Profile

ME/CFS symptom heterogeneity may partly reflect the overlay of multiple genetic predispositions. Polygenic risk scores from external GWAS (chronic pain, cognitive performance, blood pressure, iron/ferritin levels, vitamin D) applied to DecodeME participants could identify which symptom dimensions are genetically shared with other traits versus ME/CFS-specific. A patient with high ME/CFS PRS plus high pain PRS might present with a fibromyalgia-like phenotype; high ME/CFS PRS plus high cognitive-performance PRS (protective direction) might have preserved cognition despite severe PEM.

(Certainty: 0.50)

Falsifiable prediction: If no external PRS predicts any ME/CFS symptom dimension (all $p > 0.01$), within-ME/CFS symptom variation is not genetically correlated with these common traits. The prediction is that pain PRS will predict pain-dominant phenotype (OR $\\approx$ 1.1), cognitive PRS will inversely predict cognitive dysfunction severity, and blood pressure PRS will predict orthostatic intolerance severity.

6.1.13 PheWAS and Pleiotropy Mapping

Open Question: Phenome-Wide Association of DecodeME Loci

For each of the eight genome-wide significant loci, a phenome-wide association study (PheWAS) using UK Biobank summary statistics can classify loci as ME/CFS-specific (no associations with correlated traits), broadly pleiotropic, or selectively pleiotropic (associated with specific trait clusters). Colocalization analysis (coloc) at each locus tests whether the same causal variant drives ME/CFS and the correlated trait. A locus that colocalises between ME/CFS and IBS but not depression identifies a gut-brain-specific variant. A locus that colocalises with depression and sleeping-too-much but not IBS identifies a central fatigue variant. The prediction—consistent with the “no shared causal variants” finding—is that two to three of eight loci will colocalise with IBS or sleep traits (posterior probability of shared variant $> 0.7$), while none will colocalise with depression.

6.1.14 Onset Age as Quantitative Trait

Open Question: GWAS of Age-at-Onset Within ME/CFS Cases

A within-case GWAS treating onset age as a continuous quantitative trait (linear regression, genotype predicting onset age, adjusted for sex and ancestry PCs) asks: which SNPs influence when ME/CFS develops, given that it develops? This is distinct from the already-proposed onset-age-stratified GWAS (which compares early versus late against controls). Onset-age modifier loci have been identified in breast cancer, Parkinson’s disease, and diabetes. In ME/CFS, with bimodally distributed onset age ($\\approx$ 18.8 and $\\approx$ 40.1 peaks), the analysis could identify developmental vulnerability windows (neurodevelopmental genes predicting early onset), aging-related pathways (mitochondrial quality control genes predicting late onset), or immune maturation variants (HLA region predicting onset relative to EBV exposure window). Onset-age modifier loci are clinically actionable: they identify which young people are at highest risk and might benefit from early intervention after infectious mononucleosis.

References

Asch, Ruth H., Chadi G. Abdallah, Richard E. Carson, and Irina Esterlis. 2024. “Challenges and Rewards of in Vivo Synaptic Density Imaging, and Its Application to the Study of Depression.” Neuropsychopharmacology 50 (1): 153–63. https://doi.org/10.1038/s41386-024-01913-3.

Birch, C. L., B. M. Wilk, M. Gajapathy, A. Brown, T. Cox, P. A. Wilmarth, D. W. Fardo, and S. Estus. 2025. “Uncovering the Genetic Architecture of ME/CFS: A Precision Approach Reveals Impact of Rare Monogenic Variation.” Journal of Translational Medicine 23: 1345. https://doi.org/10.1186/s12967-025-07586-w.

Bright, U., S. Beck, D. F. Levey, J. D. Deak, J. M. Gaziano, M. B. Stein, and J. Gelernter. 2026. “The Genetics of Fibromyalgia and Its Relationships to Psychiatric and Medical Traits.” Nature Communications 17 (1): 6248. https://doi.org/10.1038/s41467-026-75256-6.

Buchwald, Dedra, Richard Herrell, Suzanne Ashton, Michael Belcourt, Karen Schmaling, Patrick Sullivan, Michael Neale, and Jack Goldberg. 2001. “A Twin Study of Chronic Fatigue.” Psychosomatic Medicine 63 (6): 936–43. https://doi.org/10.1097/00006842-200111000-00012.

Carlo-Stella, N, S Bozzini, A De Silvestri, I Sbarsi, C Pizzochero, L Lorusso, M Martinetti, and M Cuccia. 2009. “Molecular Study of Defective Apoptosis in Myalgic Encephalomyelitis/Chronic Fatigue Syndrome.” International Journal of Immunopathology and Pharmacology 22 (4): 1057–68. https://doi.org/10.1177/039463200902200421.

Das, S., K. Taylor, J. Kozubek, J. Sardell, and A. Gardner. 2022. “Genetic Risk Factors for ME/CFS Identified Using Combinatorial Analysis.” Journal of Translational Medicine 20: 598. https://doi.org/10.1186/s12967-022-03815-8.

DecodeME Consortium, Chris P Ponting, et al. 2025. “Initial Findings from the DecodeME Genome-Wide Association Study of Myalgic Encephalomyelitis/Chronic Fatigue Syndrome.” medRxiv. https://doi.org/10.1101/2025.08.06.25333109v1.

Duncan, Laramie E, Tony Li, Madeleine Salem, Will Li, Leili Mortazavi, Hazal Senturk, Narges Shahverdizadeh, et al. 2025. “Mapping the Cellular Etiology of Schizophrenia and Complex Brain Phenotypes.” Nature Neuroscience 28 (2): 248–58. https://doi.org/10.1038/s41593-024-01834-w.

Finucane, Hilary K, Yakir A Reshef, Verneri Anttila, Kamil Slowikowski, Alexander Gusev, Andrea Byrnes, Steven Gazal, et al. 2018. “Heritability Enrichment of Specifically Expressed Genes Identifies Disease-Relevant Tissues and Cell Types.” Nature Genetics 50 (4): 621–29. https://doi.org/10.1038/s41588-018-0081-4.

Garcia, Geoffrey, Sylvie Labrouche-Colomer, Alexandre Duvignaud, et al. 2024. “Impaired Balance Between Neutrophil Extracellular Trap Formation and Degradation by DNases in COVID-19 Disease.” Journal of Translational Medicine 22: 246. https://doi.org/10.1186/s12967-024-05044-7.

Gardner, Steve. 2026. “GLP-1 RAs: Hype, Hope and Hidden Dangers.” https://investinme.org/brmec15-stevegardner.shtml.

Goebel, A., K. Busch, R. Riedl, and C. Sommer. 2021. “Passive Transfer of Fibromyalgia IgG Induces Pain and Fatigue in Mice.” Pain 162 (4): 893–904. https://doi.org/10.1097/j.pain.0000000000002134.

Goertzel, Ben N, Cassio Pennachin, Lucio de Souza Coelho, Brian Gurbaxani, Elizabeth M Maloney, and James F Jones. 2006. “Combinations of Single Nucleotide Polymorphisms in Neuroendocrine Effector and Receptor Genes Predict Chronic Fatigue Syndrome.” Pharmacogenomics 7 (3): 475–83. https://doi.org/10.2217/14622416.7.3.475.

Groven, Nina, Solveig Klæbo Reitan, Egil A Fors, and Ismail Cuneyt Guzey. 2021. “Kynurenine Metabolites and Ratios Differ Between Chronic Fatigue Syndrome, Fibromyalgia, and Healthy Controls.” Psychoneuroendocrinology 131: 105287. https://doi.org/10.1016/j.psyneuen.2021.105287.

Hu, Y., G. Yang, Z. Deng, S. Yang, Y. Li, W. Xiao, B. Lu, and X. Zhu. 2025. “The Casual Associations Between Brain Functional Networks and Fibromyalgia: A Large-Scale Genetic Correlation and Mendelian Randomization Study.” Bioengineering 12 (7): 692. https://doi.org/10.3390/bioengineering12070692.

Johnston, K. J. A., R. Signer, and L. M. Huckins. 2025. “Chronic Overlapping Pain Conditions and Nociplastic Pain.” HGG Advances 6 (1): 100381. https://doi.org/10.1016/j.xhgg.2024.100381.

Kavyani, Zahra, Ekua W Brenu, Donald R Staines, and Sonya M Marshall-Gradisnik. 2022. “Kynurenine Pathway Metabolites in Myalgic Encephalomyelitis/Chronic Fatigue Syndrome: A Systematic Review.” Journal of Translational Medicine 20 (1): 512. https://doi.org/10.1186/s12967-022-03712-3.

Kendler, K. S., J. G. M. Rosmalen, H. Ohlsson, J. Sundquist, and K. Sundquist. 2023. “A Distinctive Profile of Family Genetic Risk Scores in a Swedish National Sample of Cases of Fibromyalgia, Irritable Bowel Syndrome, and Chronic Fatigue Syndrome Compared to Migraine and Anxiety.” Psychological Medicine 53 (7): 3027–36. https://doi.org/10.1017/S0033291722000526.

Kerrebijn, Isabel, Gyda Bjornsdottir, Keon Arbabi, Lea Urpa, Hele Haapaniemi, Gudmar Thorleifsson, Lilja Stefansdottir, et al. 2026. “The Genetic Architecture of Fibromyalgia Across 2.5 Million Individuals.” Nature Medicine 32 (8): 3060–70. https://doi.org/10.1038/s41591-026-04492-6.

King, Eric A, J Wade Davis, and Jacob F Degner. 2024. “Are Drug Targets with Genetic Support Twice as Likely to Be Approved? Revised Estimates of the Impact of Genetic Support for Drug Mechanisms on the Probability of Drug Approval.” Nature. https://doi.org/10.1038/s41586-024-07316-0.

Komaroff, Anthony L, and W Ian Lipkin. 2023. “ME/CFS and Long COVID Share Similar Symptoms and Biological Abnormalities: Road Map to the Literature.” Frontiers in Medicine 10: 1187163. https://doi.org/10.3389/fmed.2023.1187163.

Koopmans, Frank, Pim van Nierop, Maria Andres-Alonso, Andrea Byrnes, Tony Cijsouw, Marcelo P. Coba, L. Niels Cornelisse, et al. 2019. “SynGO: An Evidence-Based, Expert-Curated Knowledge Base for the Synapse.” Neuron 103 (2): 217–234.e4. https://doi.org/10.1016/j.neuron.2019.05.002.

Lee, Jun Hyun et al. 2026. “Global and Local Genetic Overlap Among ME/CFS, Irritable Bowel Syndrome, and Psychiatric Traits: A Hypothesis-Generating Analysis.” medRxiv. https://doi.org/10.64898/2026.06.08.26355171v1.

Liao, Y, JG Qi, H Yan, QY Zhang, TY Ji, XZ Chang, HP Yang, HF Jin, and JB Du. 2021. “Comorbidity of Chronic Fatigue Syndrome, Postural Tachycardia Syndrome, and Narcolepsy with 5,10-methylenetetrahydrofolate Reductase (MTHFR) Mutation in an Adolescent: A Case Report.” Chinese Medical Journal 134 (12): 1495–97. https://doi.org/10.1097/CM9.0000000000001387.

Lin, L., Y. Li, F. Ji, J. Lin, M. Zhu, D. Liang, M. Cao, G. Fu, and Y. Fu. 2026. “Shared Latent Genetic Liability Across Fibromyalgia and Psychiatric Traits: Novel Insights from Genomic Structural Equation Modeling.” PLoS Genetics 22 (1): e1012034. https://doi.org/10.1371/journal.pgen.1012034.

Maccallini, P. 2026. “Biological Insights from Genome-Wide Association Studies and Whole Genome Sequencing of Myalgic Encephalomyelitis/ Chronic Fatigue Syndrome.” Research Square [Preprint], June. https://doi.org/10.21203/rs.3.rs-9702020/v1.

Marshall-Gradisnik, Sonya, Tahnee Huth, Anu Chacko, Samantha Johnston, Peter Smith, and Donald Staines. 2016. “Natural Killer Cells and Single Nucleotide Polymorphisms of Specific Ion Channels and Receptor Genes in Myalgic Encephalomyelitis/Chronic Fatigue Syndrome.” Applied Clinical Genetics 9: 39–47. https://doi.org/10.2147/TACG.S99405.

ME/CFS Science. 2025a. “DecodeME: The Biggest ME/CFS Study Ever.” 2025. https://mecfsscience.org/decodeme-the-biggest-me-cfs-study-ever/.

———. 2025b. “Genes Pointing to the Brain: DecodeME Part II.” 2025. https://mecfsscience.org/genes-pointing-to-the-brain-decodeme-part-ii/.

Natelson, Benjamin H, Jin-Mann S Lin, Michelle Blate, Sarah Khan, Yang Chen, and Elizabeth R Unger. 2022. “Physiological Assessment of Orthostatic Intolerance in Chronic Fatigue Syndrome.” Journal of Translational Medicine 20 (1): 95. https://doi.org/10.1186/s12967-022-03289-8.

Onwordi, Ellis C., Els F. Halff, Thomas Whitehurst, Ayla Mansur, Marie-Caroline Cotel, Lisa Wells, Hannah Creeney, et al. 2020. “Synaptic Density Marker SV2A Is Reduced in Schizophrenia Patients and Unaffected by Antipsychotics in Rats.” Nature Communications 11 (1): 246. https://doi.org/10.1038/s41467-019-14122-0.

Onwordi, Ellis C., Thomas Whitehurst, Ayla Mansur, Ben Statton, Alex Berry, Marina Quinlan, Declan P. O’Regan, et al. 2021. “The Relationship Between Synaptic Density Marker SV2A, Glutamate and n-Acetyl Aspartate Levels in Healthy Volunteers and Schizophrenia: A Multimodal PET and Magnetic Resonance Spectroscopy Brain Imaging Study.” Translational Psychiatry 11 (1): 393. https://doi.org/10.1038/s41398-021-01515-3.

Regland, Bjorn, Sara Forsmark, Lena Halaouate, Michael Matousek, Birgitta Peilot, Olof Zachrisson, and Carl-Gerhard Gottfries. 2015. “Response to Vitamin B12 and Folic Acid in Myalgic Encephalomyelitis and Fibromyalgia.” PLoS ONE 10 (4): e0124648. https://doi.org/10.1371/journal.pone.0124648.

Sardell, Jessica M., S. Das, M. Pearson, J. Kozubek, K. Taylor, and A. Gardner. 2026. “Identification of Novel Reproducible Combinatorial Genetic Risk Factors for Myalgic Encephalomyelitis in the DecodeME Patient Cohort and Commonalities.” Journal of Translational Medicine 24: 420. https://doi.org/10.1186/s12967-026-08167-1.

Serrano, Maria-Eleni, Eugene Kim, Marija M. Petrinovic, Federico Turkheimer, and Diana Cash. 2022. “Imaging Synaptic Density: The Next Holy Grail of Neuroscience?” Frontiers in Neuroscience 16: 796129. https://doi.org/10.3389/fnins.2022.796129.

Snyder, Mark P, Hongjuan Zhao, et al. 2025. “Rare Coding Variants in ME/CFS Implicate Neuronal Genes and Synaptic Function.” medRxiv. https://doi.org/10.1101/2025.04.15.25325899v2.

Steen, O. D., H. Ohlsson, S. L. van Ockenburg, K. S. Kendler, J. Sundquist, and K. Sundquist. 2026. “Shared Genetic Risk Between Functional Somatic Syndromes, Internalizing Disorders, and Immune-Mediated Diseases: A Twin-Sibling Study.” Brain, Behavior, and Immunity 117: 1–10. https://doi.org/10.1016/j.bbi.2026.106837.

Syed, Abu Mohammad, Alexander K Karius, Jin Ma, Ping-yuan Wang, and Paul M Hwang. 2025. “Mitochondrial Dysfunction in Myalgic Encephalomyelitis/Chronic Fatigue Syndrome.” Physiology 40 (4). https://doi.org/10.1152/physiol.00056.2024.

Thapaliya, Kiran, Sonya Marshall-Gradisnik, Natalie Eaton-Fitch, Zahra Eftekhari, Maira Inderyas, and Leighton Barnden. 2025. “Imbalanced Brain Neurochemicals in Long COVID and ME/CFS: A Preliminary Study Using MRI.” The American Journal of Medicine 138 (3): 567–574.e1. https://doi.org/10.1016/j.amjmed.2024.04.007.

Ueland, M., R. Hajdarevic, O. Mella, Ø. Fluge, B. A. Lie, and M. K. Viken. 2022. “No Replication of Previously Reported Association with Genetic Variants in the T Cell Receptor Alpha (TRA) Locus for Myalgic Encephalomyelitis/Chronic Fatigue Syndrome.” Translational Psychiatry 12: 277. https://doi.org/10.1038/s41398-022-02046-1.

Walsh, Cara M, Nur Z Zainal, Sam J Middleton, and Eugene S Paykel. 2001. “A Family History Study of Chronic Fatigue Syndrome.” British Journal of Psychiatry 178 (5): 461–66. https://doi.org/10.1192/bjp.178.5.461.

Wirth, Klaus J., and Carmen Scheibenbogen. 2026. “Imbalance of Excitatory and Inhibitory Neurotransmitter Systems in Myalgic Encephalomyelitis/Chronic Fatigue Syndrome.” International Journal of Molecular Sciences 27 (9): 4041. https://doi.org/10.3390/ijms27094041.

Zarembska, Emilia, Klaudia Ślusarczyk, and Małgorzata Wrzosek. 2023. “The Implication of a Polymorphism in the Methylenetetrahydrofolate Reductase Gene in Homocysteine Metabolism and Related Civilisation Diseases.” International Journal of Molecular Sciences 25 (1): 193. https://doi.org/10.3390/ijms25010193.